Mad Scientist

Robotics, Artificial Intelligence & Autonomy:

Visioning Multi-Domain Warfare in 2030-2050
Technical Report
19 May 2017

DISTRIBUTION RESTRICTION: Approved for Public Release;

distribution unlimited.
This Page Intentionally Left Blank

2
 Mad Scientist
Robotics, Artificial Intelligence & Autonomy:
Visioning Multi-Domain Warfare in 2030-2050

Executive Summary
In March of 2017 the TRADOC G-2 and the Georgia Tech Research Institute (GTRI)
cosponsored a Mad Scientist Conference entitled Robotics, Artificial Intelligence &
Autonomy: Visioning Multi-Domain Warfare 2030-2050. These closely related,
interdependent technologies (robotics, artificial Intelligence, and autonomy) will exercise
key roles in future military operations, including land operations. They are being
aggressively explored and exploited by both the economies and militaries of entities
ranging from great power nation-states to super-empowered individuals. They are at
the core of the DoDs Third Offset Strategy. Although some might project a Cambrian
Explosion of transformative capabilities and applications, because DoD controls a very
small and decreasing portion of the research and development associated with these
technologies, there is real potential for a Cambrian Conundrum wherein defense
planners and strategists confront unanticipated, high order consequences from external
factors outside military or even national control.
Secure our Future is the Nations ultimate mission order to the Army, so application
of the Army philosophy of mission command particularly its components of
understand, visualize, describe, and direct were applied to organize the insights of
this Mad Scientist project.
Understand. This Technical Report first endeavors to understand the relevant
trends for robotics, artificial intelligence and autonomy. Each of these technology
areas is explored to establish their definition and a broad understanding of their
baseline state-of-the art. The Report captures Mad Scientist projections for each
technology, together with key projected challenges for their further development, and
estimates of their relevance for future military operations. The Report then further
reinforces understanding of the relevant trends by describing representative threat
developments with respect to these technologies, as well as perhaps the most relevant
trend of all: the extraordinary speed, scope and convergence of these technologies.
Visualize. TRADOC has identified five key Future
Operating Environment (FOE) characteristics. FOE Characteristics
This Technical Report next undertakes to
visualize the potential of five potential solution o Contested in all Domains
approaches that address the characteristics of o Unprecedented Speed
the Future Operating Environment. The five o WMD Proliferation
o Complex Terrain the Norm
solution approaches emerged during Mad Scientist
o Hybrid Combatants
discussions and included

3
 Manned-Unmanned Teaming (MUM-T),
Asymmetric Awareness & Decision,
Swarming,
Intelligent Networks for the Internet of Battle Things,
Autonomous Sustainment.

This Mad Scientist project conducted a concurrent SciTech Futures Crowd-sourcing

wargame to connect disparate thought leaders in an exploration of how robotics,
artificial intelligence, and autonomous systems and related technologies might
transform the world, and the implications of that transformation for the Army. Summary
tables of the crowd-sourcing ideas relevant to each solution approach are included in
the visualize section (III), but
described in greater detail at Competitions of Multi-Domain Warfare
Appendix B to this report.
o Finders vs Hiders
Describe. Although the Army o Strikers vs Shielders
continues to explore the issues of o Range & Lethality vs
multi-domain warfare, it is already o Close Engagement & Survivability
possible to describe the o Disconnection / Disaggregation / Decentralization
competitions of multi-domain vs Connection / Aggregation / Centralization
warfare and how the solution o Offense vs Defense
approaches previously visualized o Planning & Judgement vs Reaction & Autonomy
o Escalation vs De-Escalation
can be applied to those competitions
o Domain vs Domain
Direct. As future competitors o Dimension vs Dimension
leverage the technologies of
robotics, artificial intelligence, and autonomy in the competitions of multi-domain
warfare, success will accrue to those competitors most successful in the institutional
contests already underway that will shape the outcomes of the future. The final
section of this Technical Report addresses ways the Army can direct the drivers of
outcome
Strategy and Policy
Concepts
Innovation & Adaptation
Combinations
Learning

This Mad Scientist project to understand, visualize, describe and direct the dynamic and
converging fields of robotics, artificial intelligence and autonomy will not only enable
mission command but also help set conditions for future adaptation and operational
success in multi-domain warfare.

4
 Table of Contents

Executive Summary ................................................................................................................................... 3

I Introduction: Robotics, Artificial Intelligence & Autonomy: Visioning Multi-Domain Warfare
2030-2050 ................................................................................................................................................... 7
Mad Scientist Conference: Robotics, Artificial Intelligence & Autonomy: Visioning Multi-
Domain Warfare 2030-2050 ................................................................................................................. 8
Mad Scientist SciTech Crowd-Sourcing Exercise ............................................................................. 9
Robotics, Artificial Intelligence and Autonomy: The Coming Cambrian Conundrum .............. 10
II UNDERSTAND: Relevant Trends for Robotics, Artificial Intelligence and Autonomy ............... 15
Autonomy Trends ................................................................................................................................. 15
Artificial Intelligence (AI) Trends ........................................................................................................ 18
Robotic Trends ..................................................................................................................................... 22
Threat Trends ....................................................................................................................................... 25
Technology Trends: Speed, Scope and Convergence .................................................................. 27
III VISUALIZE: Solution Approaches that address the Characteristics of the Future Operational
Environment .............................................................................................................................................. 31
Overview: the Characteristics of the Future Operational Environment ........................................ 31
Manned Unmanned Teaming (MUM-T) ............................................................................................ 32
Asymmetric Awareness and Decision (AA&D) ................................................................................ 34
Swarming ............................................................................................................................................... 37
Intelligent Networks for the Internet of Battle Things (IoBT) ....................................................... 38
Autonomous Sustainment ................................................................................................................... 40
IV DESCRIBE: the Competitions of Multi-Domain Warfare ............................................................... 43
Overview: the Competitions of Multi-Domain Warfare ................................................................... 43
Finders vs Hiders ................................................................................................................................. 44
Strikers vs Shielders ............................................................................................................................ 44
Range & Lethality vs Close Engagement & Survivability ............................................................... 45
Disconnection / Disaggregation / Decentralization vs Connection / Aggregation /
Centralization ........................................................................................................................................ 46
Offense vs Defense ............................................................................................................................. 47
Planning & Judgment vs Reaction & Autonomy .............................................................................. 48
Escalation vs De-Escalation ............................................................................................................... 49

5
 Domain vs Domain ............................................................................................................................... 49
Dimension vs Dimension..................................................................................................................... 50
V DIRECT: the Drivers of Outcome...................................................................................................... 51
Strategies & Policy ............................................................................................................................... 51
Concepts................................................................................................................................................ 53
Innovation & Adaptation ...................................................................................................................... 55
Combinations ........................................................................................................................................ 58
Learning ................................................................................................................................................. 59
VI Conclusion ............................................................................................................................................ 63
End Notes .................................................................................................................................................. 64
Appendix A: Workshop Design & Sources ........................................................................................... 71
Appendix A-1: Workshop Agenda ......................................................................................................... 71
Appendix A-2: Conference Presenters ............................................................................................. 75
Appendix A-3: Conference Presentations ........................................................................................ 77
Appendix B: SciTech Crowd-Sourcing Insights ................................................................................... 79
Appendix C: Collection and Assessment Methodology...................................................................... 87
Appendix D: References ......................................................................................................................... 95

6
I Introduction: Robotics, Artificial Intelligence & Autonomy:
Visioning Multi-Domain Warfare 2030-2050

Mad Scientist (MS) is a Training and Doctrine Command G-2 (Intelligence) initiative that
explores a series of future Army challenges through an open, public dialogue with a
broad range of Joint, interagency and international partners; academia; policy
institutions; and the private sector. Mad Scientist events are part of the G-2s
continuous study of the future Operational Environment out to 2050, as well as the Army
Capabilities Integration Center (ARCIC) Campaign of Learning and 2025 Maneuvers.
In March of 2017 the TRADOC G-2 and the Georgia Tech Research Institute (GTRI)
cosponsored a Mad Scientist* Conference entitled Robotics, Artificial Intelligence &
Autonomy: Visioning Multi-Domain Warfare 2030-2050. These closely related,
interdependent technologies (robotics, artificial Intelligence, and autonomy) will exercise
key roles in future military operations, including We are on the cusp of a variety of
land operations. They are being aggressively breakthroughs that will be as profound as
explored and exploited by both the economies the internal combustion engine and
and militaries of entities ranging from great machine gun was on combat circa WWI.
power nation-states to super-empowered
August Cole, Mad Scientist
individuals. They are at the core of the DoDs Conference, 7 Mar 2017
Third Offset Strategy.
Conference participants shared a wide range of views with respect to the current state
of these technologies, developmental challenges and areas of future research, and of
course their potential applications in both military and non-military endeavors. They
were guided by two key questions:
How can Artificial Intelligence (AI), and autonomy effectively support regional,
global, Joint, and Army operations in Multi-Domain Warfare, 2030-2050, as well
as those capabilities a potential adversary may employ?
How may AI and robotics change the relationship between humans and warfare;
what insights will contribute to a greater understanding of conflict and the
character of war in the Future Operating Environment?
This Mad Scientist project addresses technologies that although already enjoying
extensive application in our daily lives have only traversed a small fraction of their

* For the remainder of this Technical Report, the term Mad Scientist will
connote any Mad Scientist conference presenter, participant, or crowd-
sourcing exercise contributor for the Mad Scientist Robotics, Artificial
Intelligence, and Autonomy project.

projected growth paths. Accurately assessing those growth paths out to 2050 is

7
daunting indeed. However, effective foresight the process of thinking about our world
and how it might change is critical to yielding better judgments about how to best
prepare for whatever the future may bring.1 It is the intent of this study to paint a picture
of key issues for the Army at the intersection of these emerging technologies and
landpower, thereby assisting Army leaders in exploring the key decisions and actions
needed to defend the Nation in the Future Operating Environment.

Mad Scientist Conference: Robotics, Artificial Intelligence &

Autonomy: Visioning Multi-Domain Warfare 2030-2050

The Mad Scientist Conference was held in co-sponsorship with the Georgia Tech
Research Institute (GTRI) in Atlanta, Georgia on 7-8 March 2017. The conference
included 18 presentations and one panel of three members. Participants included LTG
Kevin Mangum (DGG, TRADOC); MG Robert M. Dyess (Deputy Director, Army
Capabilities Integration Center); Dr. Steve Cross, Executive Vice President for
Research, Georgia Institute of Technology; Dr. Augustus Fountain, Deputy Chief
Scientist (ST) Office of the Deputy Assistant Secretary of the Army (Research &
Technology); and Dr. Robert Sadowski, U.S. Army Chief Roboticist, Robotics Senior
Research Scientist, and Research, Technology and Integration Director at U.S. Army
TARDEC. Conference presentations are listed at Appendix A-3 and are accessible at
the following link: https://community.apan.org/wg/tradoc-g2/mad-scientist/p/robotics_ai
Notes from speaker presentations and panel discussions are synthesized into this
Technical Report.

This Mad Scientist event is the most recent of a series. Others over the last several
months have included:

Disruptive Technologies. Co-hosted by Georgetown University, addressed

sentient data, internet of sustainable energy, platform mergers, autonomous vs
unmanned systems, and the next revolution in computing
Human Dimension. Co-hosted by Army University, explored measuring
cognitive potential, man-machine interface, genome sequencing, wearables,
continuous diagnostics, and performance enhancers
Megacities and Dense Urban Areas. Co-hosted by Arizona State University,
explored the modeling of megacities, population-centric intelligence, invisible
geography, hot zone robotics, avatars in the field, and the role of augmented and
virtual reality in training for operations in dense urban areas.
Strategic Security Environment in 2025 and Beyond. Co-hosted by
Georgetown University, explored the thesis that the direction of global trends
shaping the future Operational Environment (2030-2050), and the geopolitical
situation that results from it, will fundamentally change the character of warfare.
The 2050 Cyber Army. Co-hosted by the Army Cyber Institute at the United
States Military Academy, visualized the Armys Cyber Force in 2050. Although
this Mad Scientist project encompassed a wide range of cyber domain topics, its

8
 focus was to better understand what the Army may need to do to build the cyber
workforce and develop partnerships in order to address DoD missions in
cyberspace in the 2050 time frame.2

In addition, the analysts drew on multiple sources relevant to the event topics, including:

Joint Concept for Robotics and Autonomous Systems (JCRAS) (19 October
2016).
U.S. Army Robotic and Autonomous Systems Strategy (RAS) (January
2017).
Defense Science Board Summer Study on Autonomy (June 2016).
Center for Naval Analyses Andrew Ilachinski Study AI, Robots, and Swarms:
Issues, Questions, and Recommended Studies (January 2017).
Draft TRADOC Paper: The Operational Environment, 2035-2050: The
Emerging Character of Warfare.
Other references as cited in Appendix E to this report.

Mad Scientist SciTech Crowd-Sourcing Exercise

The SciTech Futures Project is a partnership between the Deputy Assistant Secretary of
the Army (Research & Technology) and the USC Institute for Creative Technologies
(ICT), one of many US Government efforts aimed at leveraging the collective wisdom
and ability of the American public. During the period from 6 to 19 March 2017 it
conducted a futures game, sponsored by the Office of the Deputy Assistant Secretary of
the Army (Research & Technology) in partnership with the U.S. Army Training and
Doctrine Commands (TRADOC) Mad Scientist Initiative. The crowd-sourcing exercise
sought to connect disparate thought leaders in exploring how advances in robotics,
artificial intelligence, autonomous systems, and related technologies might transform
the world - and the implications of that transformation for the Army. Participants
leveraged an interactive web site to share their ideas about the future, collaborate with
(and challenge) other players, and bid on the most compelling concepts in an online
marketplace. The output of the SciTech Futures Crowd-sourcing Exercise is described
in more detail at Appendix B.

9
 FIG I-1 SciTech Futures Crowd-Sourcing Project: https://scitechfutures.com/ex6/workshop/

Robotics, Artificial Intelligence and Autonomy: The Coming

Cambrian Conundrum

Although the burgeoning impact of robotics and autonomy has been evident for some
time, we are now recognizing the rapidly accelerating emergence of robotics, artificial
intelligence, and autonomy in our daily lives. Fossil records demonstrate the sudden
appearance about 542 million years ago -- of complex animals with mineralized
skeletal remains. Some describe this Cambrian Explosion 3 as the most significant
event in Earths evolutionary history, one that irreversibly changed the biosphere and
led to a stunning diversity of body forms and types.4 Today, many surmise that the
impacts of robotics, artificial intelligence and autonomy together with their derivatives,
e.g., machine learning -- are about to induce a metaphorical Cambrian Explosion of
transformative capabilities and applications. For the Department of Defense, however,
which controls a minute and ever-shrinking portion of the research in these fields, this
Cambrian Explosion is more probably a Cambrian Conundrum. Defense planners and
strategists will confront unanticipated, high-order consequences from external factors
principally outside military or even national control.
Many scientists believe that the original Cambrian Explosions transformative
evolutionary developments were triggered by a complex interplay of relatively small
environmental changes,5 and the emergence of better eyes, nervous networks, and the

10
ability to move and interact with the world. A similarly complex interplay is clearly at
work across the topics of robotics, artificial intelligence and autonomy as sensors,
actuators, and processors get both
cheaper and better. The fields of While difficult to quantify, the study concluded that
robotics, artificial intelligence and autonomyfueled by advances in artificial
autonomy share a number of intelligencehas attained a tipping point in value.
enabling technologies, research Autonomous capabilities are increasingly ubiquitous
challenges, and future use cases; it and are readily available to allies and adversaries
is indeed difficult to discuss one in alike. The study therefore concluded that DoD must
the absence of the others. take immediate action to accelerate its exploitation
Autonomous, for example, is a of autonomy while also preparing to counter
quality of a robotic system; autonomy employed by adversaries.
autonomous swarms are typically DSB Summer Study on Autonomy
multi-robotic configurations. The June 2016
following statement in a recent
Center for Naval Analyses Report demonstrates their seamlessness and interaction: In
short, autonomous systems are inherently, and irreducibly, artificially intelligent
robots.6 This Mad Scientist project properly approaches these inter-related topics
simultaneously and comprehensively.
This Mad Scientist project is also timely: these technologies are at the core of the DoD
Third Offset Strategy, and all Mad Scientists at the March Conference noted that we
are still at the start of the learning curve for both the potential -- and the challenges --
associated with these technologies. The dialogue between technology subject matter
experts and military practitioners is both appropriate and necessary, as demonstrated
by these two excerpts from Day One Conference discussions:
Scientists are great at envisioning technology, but need the assistance of
others to understand the opportunities, challenges and pitfalls of it.7
Dr. Augustus Fountain, Deputy Chief Scientist
(ST) Office of the Deputy Assistant Secretary of
the Army (Research & Technology)

Technology will outpace experience. 30 years experience will not help

you understand technology that is 6 months old.8
Mr. Tom Greco, TRADOC G2

Analysis Approach. Secure the Future is the Nations ultimate mission order to the
Army, so application of the Army philosophy of mission command particularly its
components of understand, visualize, describe, and direct is a useful organizational
rubric for the insights of this Mad Scientist project.

11
The report first endeavors to understand the trends, threats and opportunities
associated with these technologies, together with the speed, scope and convergence of
their technology impacts.

Understand: Relevant Trends for Robotics, Artificial Intelligence and

Autonomy
Autonomy Trends
Artificial Intelligence Trends
Robotics Trends
Threat Trends
Trend Speed, Scope and Convergence

The report then proceeds to visualize how the incorporation of this understanding into
five potential solution approaches dominant and integrating themes throughout the
Mad Scientist discussions -- might address the characteristics of the emerging Future
Operational Environment.

Visualize: Solution Approaches that address the Characteristics of the Future

Operational Environment
Manned-Unmanned Teaming (MUM-T)
Asymmetric Awareness & Decision
Swarming
Intelligent Networks for the Internet of Battle Things
Autonomous Sustainment
The future will feature the interaction of multiple great powers similarly equipped with
emerging technologies including robotics, artificial intelligence and autonomy -- and
simultaneously trying to address the strategic and operational challenges of the future
operational environment. This interaction will drive a fundamental change in the
character of warfare, a change characterized as a series of competitions.9 The next
section describes the relevance of these solution approaches to those competitions.

Describe: The Competitions of Multi-Domain Warfare

Finders vs Hiders
Strikers vs Shielders
Range & Lethality vs Close Engagement & Survivability
Disconnection / Disaggregation / Decentralization
vs Connection / Aggregation / Centralization
Offense vs Defense
Planning & Judgement vs Reaction & Autonomy
Escalation vs De-Escalation
Domain vs Domain
Dimension vs Dimension

12
Finally, the report summarizes how the United States Army might direct the drivers that
will shape the outcome of these future competitions.

Direct: The Drivers of Outcome

Strategy & Policy
Concepts
Innovation & Adaptation
Combinations
Learning

The actions we take today with respect to these key outcome drivers will shape our
future success in leveraging these technologies and mitigating their risk.

13
This Page Intentionally Left Blank

14
II UNDERSTAND: Relevant Trends for Robotics, Artificial
Intelligence and Autonomy
To understand the potential impact of robotics, artificial intelligence, and autonomy we
must first review the definitions, baseline, projected trends, challenges and fundamental
relevance for each. This review will also address these trends from a threat
perspective, assess the rate and scope of technological progress, and illustrate how
these technologies interact and converge.

Understand: Relevant Trends for Robotics, Artificial Intelligence and

Autonomy
Autonomy Trends
Artificial Intelligence Trends
Robotics Trends
Threat Trends
Trend Speed, Scope and Convergence

Autonomy Trends

Autonomy Definition. The Joint Concept for Robotics and Autonomous Systems
defines autonomy as follows:

the level of independence that humans grant a system to execute a given task. It is
the condition or quality of being self-governing to achieve an assigned task based on the
systems own situational awareness (integrated sensing, perceiving, analyzing), planning
and decision-making. Autonomy refers to a spectrum of automation in which independent
decision-making can be tailored for a specific mission, level of risk, and degree of human-
machine teaming.10

The phrase spectrum of automation alludes to the fact that there are different degrees
to autonomy, identified by Mad Scientists as:11

Fully Autonomous: Human Out of the Loop: no ability for human to

intervene in real time.
Supervised Autonomous: Human on the Loop: humans can intervene
in real time.
Semi-Autonomous: Human in the Loop: machines wait for human input
before taking action.
Non-Autonomous (Remote Control): human in the loop via remote
controls; no autonomy in system.12

Mad Scientists noted that there are also three unique dimensions to the idea of
autonomy and that these dimensions constitute fundamentally distinct concepts that are
problematically applied to the same word.13 First, there is the autonomy dimension of
the human-machine command and control relationship described above in the

15
degrees of autonomy. Next, there is a dimension that addresses the innate
sophistication of the machine, a sophistication manifested in a range that includes:

Automatic: Simple, threshold based

Automated: Complex, Rule-Based
Autonomous: Self-learning / evolving
Intelligent: Human level cognition of a problem

As sophistication increases, autonomous systems are paradoxically both more capable

but also less explainable. The final, and most critical dimension to the idea of autonomy
is the complexity of the task performed. In the words of one Mad Scientist, Both a
landmine and a toaster are automatic systems. A system is autonomous with respect to
what task?14 No system is fully autonomous with respect to all tasks.

Autonomy Baseline. Mad Scientists agreed that autonomy is already here in many
aspects of our daily life, citing numerous examples in the fields of agriculture,
environmental monitoring, utilities management, and many others.15 Even news
coverage applies autonomy:

Social media bots, like it or not, shape the information environment. During the 2016 US
presidential campaign debates and on election day, the Oxford Internet Institute estimates
that 17 27% of all the tweet traffic was generated by highly automated accounts or
bots.

Mad Scientist SciTech Crowd-sourcing Exercise16

Autonomy is also already evident on the battlefield. At least 30 countries have

defensive, human-supervised autonomous weapons such as the Aegis and Patriot.17
The AH-64D Apache attack helicopters Longbow fire control radar already
automatically searches, detects, locates, classifies, and prioritizes multiple moving and
stationary targets on land, air, and water in all weather and battlefield conditions.18

Some fully autonomous weapon systems are also emerging. The Israeli Harpy drone
(anti-radiation loitering munition) has been sold to India, Turkey, South Korea, and
China. China reportedly has reverse-engineered their own variant. The U.S. has
experimented with similar systems in the Tacit Rainbow and the Low Cost Autonomous
Attack System (LOCAAS) programs. Although both these projects have been
cancelled, they illustrate our willingness to explore high levels of autonomy.19

Autonomy Projections. Mad Scientists expect autonomy to evolve from solutions that
are reactive, single platform, point solutions under minimal human control to solutions
that are flexible, multi-modal, and goal-oriented featuring trusted man-machine
collaboration, distributed autonomy and continuous learning.20

16
 Collaborative Autonomy will be learning and adaptation to perform a new task
based on mere demonstration of the task by end-users (factory workers, service
workers, consumers, Soldiers) to teach the robot what to do.21

Distributed Autonomy will be dynamic team formation from heterogeneous

platforms to include coordination in settings with limited or impaired
communication and the emergence of new tactics and strategies enabled by
multiagent capabilities.22

Continuous Learning
will be a continuous,
incremental evolution and
expansion of capabilities,
to include the
incorporation of high-level
guidance (such as human
instruction, changes in
laws / ROEs / constraints)
and Transfer Learning:
bootstrapping from
knowledge learned on
other tasks, in other
domains, and by different
platforms. Learning to
Learn will exploit
opportunities to learn
based on self-awareness
of current limitations.23

Projections for the transportation

industry are particularly well-
developed, as illustrated by the
Wall Street Journal article at FIG
II-1.24 Mad Scientists imagined
ubiquitous autonomous vehicles
able to drive, talk, entertain, and
even self-maintain.25 FIG II-1

Autonomy Challenges. Mad Scientists acknowledged that the future projections for
the field of autonomy simultaneously pose challenges:

Goal-Oriented Autonomy. Decision and adaptation like a human will struggle

to include the incorporation of ethics and morality into decision-making.26

Trusted Collaboration. The challenge of trust between man and machine is a

dominant theme of both the Mad Scientist observations as well as other writings
on this topic. Machines must properly perceive human goals and preserve their
17
 autonomous system integrity while achieving joint man-machine goals in a
manner explainable to and completely trusted by -- the human component.27

Distributed Systems. Rethinking the execution of tasks using multiple,

distributed agents while preserving command-level understanding and decision
adds an additional layer of complexity to the already challenging task of
designing and building autonomous systems.28

Transfer Learning: Learning by inference from similar tasks must address the
challenges of seamless adaptation to changing contexts and environments,
including the contextual inference of missing data and physical attributes.29

High Reliability Theory. Normal Accident Theory holds that, no matter what
organizations do, accidents are inevitable in complex, tightlycoupled systems.
High Reliability Theory asserts that organizations can contribute significantly to
the prevention of accidents.30 Because of the significant complexity and tight
coupling of future autonomous systems, there is an obvious challenge in the
application of high reliability theory to emerging technologies that are not yet well
comprehended.31

Relevance of Autonomous Systems. For the foreseeable future, no autonomous

system will have the breadth, robustness and flexibility of human cognition, but
autonomous systems offer the potential for speed, mass, and penetration capabilities in
future lethal, high threat environments while minimizing risks to Soldiers.

Artificial Intelligence (AI) Trends

AI Definition. Richard Potember of the Mitre Corporation offers the following definition
for AI. AI is

conventionally, if loosely, defined as intelligence exhibited by machines. Operationally,

it can be defined as those areas of R&D practiced by computer scientists who identify with
one or more of the following academic sub-disciplines: Computer Vision, Natural Language
Processing (NLP), Robotics (including Human-Robot Interactions), Search and Planning,
Multi-agent Systems, Social Media Analysis (including Crowdsourcing), and Knowledge
Representation and Reasoning (KRR). The field of Machine Learning (ML) is a
foundational basis for AI32

Mad Scientists cited numerous key components to the field of AI, including:33
Automated Perception using a range of modalities: vision, sonar, lidar, haptics;
Robotic Action such as locomotion and manipulation;
Deep Reasoning: planning, goal-oriented behavior, projection;
Language Technologies: language, speech, dialog, social nets;
Big Data: storage, processing, analytics and inference;34
Machine Learning to include adaptation, reflection, knowledge acquisition.

Mad Scientists treated AI and Cognitive Computing as interchangeable terms.35

18
AI Baseline. Physical robots are merely
one type of AI entity. Others include
cyber agents, decision aids, the internet
of things, and increasingly: munitions and
networks.36 Mad Scientists described AI
as a key component of the Fourth
Industrial Revolution. (FIG II-2).
Currently there is a $153B market for AI-
enabled technology -- with an estimated
annual creative disruption impact of $14-
33 trillion.37

AI technology is currently advancing at

breakneck speeds, with recent interesting FIG II-2: Carbonell Presentation
accomplishments in a broad range of Mad Scientist Conference Day One
areas to include:38

o Unsupervised learning, generative modeling;

o Deep Learning exploiting Deep Neural Networks (DNNs) to facilitate
automated interpretation of vision and speech (Neural Language Processing;
o Reinforcement learning for decision-making and robotics training;
o Multi-task networks, transfer learning;
o Use of simulated data;
o Large-margin methods (SVM) for entity classification;
o Graphical models.

Last week, the biggest investment firm in the world laid off a bunch of its top stock pickers and
replaced them with computer programs. This is happening all over Wall Street. Firms are moving
away from having humans decide what stocks to buy and sell and towards having humans
program computers and then letting the computers decide what to buy and sell. Computers are
cheaper than humans. They are more disciplined. They can think about more things at once. Like,
they can scan Facebook for trends, they can count the number of cars in Wal-Mart parking lots,
and then use all that to figure out what stock to buy and sell and do it automatically. This is the
way the world is going. This is what the stock market is becoming.

NPR Planet Money Podcast: BOTUS

Episode 763, April 7, 2017

AI Projections. AI touches virtually every area of computer science and in the words of
one Mad Scientist: Everything that we formerly electrified, we will now cognitize:39no
more dumb data.40 Big parts of the global economy will be run by AI, with widespread
disruption to the electrical infrastructure, healthcare, additive manufacturing,
transportation sector, supply chain management, and farming. This disruption is not

19
confined to blue-collar labor markets, it is also advancing in white-collar fields such
as financing and equity trading.41

Autonomy and learning are already pervasive in sensing, but will increasingly take over
decision-making as well.42 Mad Scientists project future AI capable of reflection,
curiosity and teamwork.43 AI may extend language translation capabilities, perhaps
even to other species.44 Individuals may eventually exercise ubiquitous personalized
agents (COGs), and artificial intelligence will extend the boundary of self. Human
judgment will remain essential, but the line of decision allocation between humans and
machines will be shifting in coming years.45

AI Challenges.

Maturity. Current AI systems are frequently brittle: narrow applications that can
generate very dumb results when operated outside of narrow constraints. They are
also vulnerable to spoofing.46

Big Data and Active Learning. Big Data is the fuel that drives deep learning,
and is big not only from a quantity perspective. It is also big from the perspective of
a high level of complexity (potential relations among entries) and dimensionality
(attributes per entry).47

Welcome to the Big Data Revolution. Data now streams from daily life: from phones and credit cards to
computers and sensor-equipped buildings. In 2013, IBM released some numbers pointing to the fact that 90% of
ALL of the worlds data has been produced in the past TWO years (and were confident to assume that it has
since grown even higher.) The exponential growth of online data can largely be attributed to the advent and
maturing of social media, analytics platforms and the ongoing move of mobile tech from analog to digital
technologies. There is a big data revolution, says Weatherhead University Professor Gary King. But it is not
the quantity of data that is revolutionary. The big data revolution is that now we can do something with the
data.

Bedrock Data Web Site

Paradoxically, Big Data is often associated with Knowledge Sparsity because only a
tiny fraction of the vast amounts of Big Data is effectively labeled. Less than .01% of all
galaxies in the Sloan Sky Survey have consensus labels; less than .0001% of all web
pages have topic labels. Less than .0001% of all financial transactions are investigated
and labeled as fraudulent / non-fraudulent. Mad Scientists described Active Learning
as a potential technique to address knowledge sparsity by teaming AI capabilities with
external assistance that selects the portions of Big Data with maximum potential impact
on learning.48 49

DoD Problem Set. Mad Scientists acknowledge that there are unique
characteristics of the DoD space including a lack of data, more complex sensing
phenomena, the requirement for multi-source fusion and distributed sensing, and the

20
high consequences of military decision-making. Current DoD acquisition processes,
moreover, cannot keep pace with the transformative rate of change in the AI field.50

Explainability. The complexity of AI systems is a double-edged sword, wherein

enhanced capability is paradoxically paired with decreased explainability.51 The nature
of machine learning particularly machine learning based on deep neural networks -- is

Deep learning, which came of age in the past two years thanks to faster processor architectures,
uses multiple layers of neural networks to intensify the trainingpatterns of patterns. As you go
deeper down the stack of neural networks, signals emerge for patterns that humans dont
consciously sense ... As professor Tommi Jaakkola explained to the MIT Technology Review, once
a neural network becomes extremely large, it has thousands of units per layer and maybe
hundreds of layers, then it becomes quite un-understandable. This can cause some trouble.

Andy Kessler, The Wall Street Journal: Bad Intelligence Behind the Wheel: Machine
learning will bring amazing innovationsand dangers and lawsuits (23 April 2017)

such that we often dont understand exactly how it works.52 The way such systems are
currently designed, moreover, such understanding is not possible. This is at the heart
of trust issues between the man-machine team.53

Therefore several Mad Scientists projected a future dichotomy between Safe AI

and AI in the Wild. Safe AI might come with guarantees, constraints, transparency,
and a universal undo button. Wild AI would approach full autonomy with
unrestricted adaptability, curiosity, and exploration and no ironclad guarantees.54
Artificial General Intelligence (AGI) might fit into this latter category and be a potential
game-changer of existential proportion.

Relevance of AI Systems. AI can both reinforce and mitigate the accelerating scope
and pace of warfare, integrating decision making across domains and enabling sub-
millisecond decisions.55 Expertise is perishable and doesnt scale: enhanced decision
making AI can restore balance to the OODA loop; complementing past investments in
Human personnel alone cannot adequately respond to the cyber threats facing the US military today,
Deputy Secretary of Defense Robert Work told Congress Wednesday. This is an area where we will not be
able to solve it with people, he told the Senate defense appropriations subcommittee Were putting
together the structure to watch our networks Work said. But he emphasized the need to develop
artificial intelligence and learning machines to push back against cyber threats because there just are
not enough people to defend our networks against all of the attack surfaces that we have.

Wilson Brissett, Air Force Magazine Daily Report 4 May 2017

Observe and Assessment with improved focus on Orientation and Deciding.56 AI

Battle Buddies may enhance a Soldiers personal Situational Awareness through
proactive intelligence gathering and analysis. 57 58 Training can be enhanced through
virtual / augmented realties.

21
AI may facilitate the visualization of combat effects in the cyber domain through
augmented reality. Some challenges, particularly data challenges, have such
magnitude that adequate numbers of people can simply not be mustered to address
them.59 AI will be essential in such instances.

Robotic Trends

Robotics Definition. The Joint Staff Concept for Robotics and Autonomous Systems
(JCRAS) defines robotics as

powered machines capable of executing a set of actions by direct human control,

computer control, or a combination of both. They are comprised minimally of a platform,
software, and a power source.60

The JCRAS goes on to note that Robotic and Autonomous Systems (RAS) is an
accepted term in academia and the science and technology (S&T) community; it
highlights the physical (robotic) and cognitive (autonomous) aspects of these systems.
For purposes of the JCRAS concept, RAS is a framework to describe systems with a
robotic element, an autonomous element, or more commonly, both. As technology
advances, there will be more robotic systems with autonomous capabilities as well as
non-robotic autonomous systems.61

Robotics, particularly advanced robotics, typically leverage both AI and autonomy and
are the physical manifestation by which we experience these trends in our daily lives.

There is a taxonomy for Robotic Systems that includes the following ranges of control: 62

o Remote Control. A mode of operation wherein the human operator, without

benefit of video or other sensory feedback, directly controls the actuators of a
UMS on a continuous basis, from off the vehicle and via a tethered or radio
linked control device using visual line of sight cues. In this mode, the UMS takes
no initiative and relies on continuous or nearly continuous input from the user.63
o Augmented Teleoperation. A mode of operation wherein the human operator
leverages video or other sensory feedback to directly control the actuators of a
UMS on a continuous basis.64
o Semi-Autonomy. The condition or quality of being partially self-governing to
achieve an assigned mission based on the systems pre-planned situational
awareness (integrated sensing, perceiving, analyzing) planning and decision-
making. This independence is a point on a spectrum that can be tailored to the
specific mission, level of acceptable risk, and degree of human-machine
teaming.65
o Full Autonomy. Full independence that humans grant a system to execute a
given task in a given environment.66

22
Robotics Baseline. DOD has already experienced an Accidental Robot Evolution,
with thousands of air and ground robots developed, deployed and employed in Iraq and
Afghanistan.67 Up to now the default perception has been robots as caged stupid
machines to do routine and dangerous work. Increasingly, however, robots are coming
out of the cages and migrating into our daily lives.68

Robotics Projection.69 Mad Scientists projected a future that features ever more
advanced human-robot collaboration, a collaboration that in turn will accelerate the
development of improved robotics through rapid machine learning, adaptive controls,
rapid algorithm development, and custom motion control systems. 70 Novel mechanisms
and high performance actuators will emerge as new construction paradigms are
merging component design to
According to a recent study by research firm Global
generate compact multi-function
Industry Analysts, annual spending worldwide on
systems that are both highly capable
military robotics will rise from $5.6 billion in 2012 to
and energy efficient.71 Mad Scientists
$7.5 billion by 2018. This growth will include everything
from bomb-clearing robots to pack robots that can carry
projected that human-robotic system
gear overland, unmanned underwater vehicles that can
interaction will include conversational
surveil the seas, and UAVs, more commonly known as assistants, intent and emotion
drones. The Teal Group, a U.S. consulting company, recognition, augmented reality, self-
speculated in 2013 that global spending on drones aware explainable systems, and multi-
military and civilian could cumulatively reach $89 modal communications.72
billion over the next 10 years.
Robotics are already beginning to
Michael Horowitz, Foreign Policy Magazine, 2014
transform production capabilities; this
The Looming Robotics Gap
process will accelerate as
collaborative robotic autonomy
enables robotic learning and adaptation by simple demonstration. Although a typical
current production line today features only 1 product per line, changeover cycles of 2
weeks, and a part cycle time of 6 seconds; future robotics-enabled production will be a
flexible configuration of 10+ products per line, nearly zero time required for changeover,
6 second cycle times and sub-millimeter precision.73

One Mad Scientist asserted a future for Self-Organizing Matter in the 2030-2050
timeframe, a future where almost every object will have some degree of self-assembly
and self-configuring capability,74 As the migration of robotics into our everyday
experiences advances, robotic appearances may change. It is not likely that they will
evolve to be ever more human in appearance, because humanoid shapes are sub-
optimal for many jobs or tasks. Robotic forms can be tailored to the task rather than the
other way around. Future robotics will be less immediately recognizable as robots and
our human terrain will morph to accommodate optimal robotic physical configurations.75

23
Robotics Challenges.

One to Many Control. Current robotic controls must extend from singular
entities to control of multi-robot systems: formations vice individual interaction. How do
address individual control of truly large robotic teams?76

Additive Metallic Manufacturing. To date the application of robotic 3D additive

manufacturing has focused on the use of resins and polymers to inexpensively generate
shapes and applications amenable to those materials. 3D printing of metal parts
requires relatively large and expensive machines, very high-powered lasers and
expensive technicians, although there are efforts underway to extend the desktop 3D
printing approach to metal manufacturing. Solving the 3D metallic manufacturing
problem would truly revolutionize manufacturing.77

Approximately 32,000 people die each year High Expectations. Humans will expect
in auto accidents. If autonomous vehicles high reliability performance from robotic
were able to cut that down to 20,000, people systems: death by robotic accident will be
would be uncomfortable with the idea of unacceptable, even for instances where more
deaths resulting from robotseven at only frequent death by human accident is already
10,000 deaths it would not be acceptable even tolerated for non-robotic systems.78
though humans cause many more.
Cognitive Trades. Robotics generate
Dr Magnus Egerstedt
risk reduction and performance enhancements,
Mad Scientist Presentation, March 2017
but trade the best cognitive computer available:
the human brain. This trade can be mitigated
by Centaur Warfighting: human-machine teaming that is not only possible but in many
cases preferable.79 Hybrid human-machine cognitive architectures may be able to
leverage the precision and reliability of automation without sacrificing the robustness
and flexibility of human intelligence.80

Destructive Disruption. One should also note the potential disruptive impact of
the robotics revolution, not only with respect to warfare but across the entire global
economy, particularly through the displacement of a substantial portion of the labor
force. The debate on the extent of that disruption and whether this disruption is
beneficial or detrimental remains undecided. Some have argued that technology has
always created more jobs than it has destroyed. They claim Robots Will Save the
Economy and cite robotics as necessary for further improvements in productivity
across a wide range of labor-intensive tasks.81 Others believe that the extent of the
robotics revolution is so fast and so radical that it will exceed the capacity of the labor
force to adapt.82 It is safe to assert that the robotics revolution will challenge even the
most adaptive societies and that those less adaptive may experience significant
destabilization.

Relevance of Robotic Systems. Robotic systems mitigate the risk of combat while
providing significant performance advantages such as speed, efficiency, and resilience.
Robotic sensor applications, for example, might include precision sensor positioning,

24
sensor placement in adverse environments, and multiple, distributed sensors and
platforms.83 Just as robotics may advance manufacturing to the next industrial
renaissance,84 they may also enable transformative efficiencies in the transportation
and sustainment of land forces.

Threat Trends

The United States is not the only competitor researching and fielding these
technologies. The threat perspective must account for nation states, violent non-state
actors (VNSAs), and according to Mad Scientists potentially the technologies
themselves.

Nation States

Russia. Russians view these robotics, artificial intelligence and autonomy

technologies as key to a next Kondratiev Sixth Wave.85 They have fielded a system
called the URAN-9, an unmanned combat vehicle with a 30 mm gun and anti-tank
guided missiles. At 9 tons it is readily air-liftable and demonstrates the advantageous
robotic trade space between reduced (or eliminated) human survivability requirements
and system size, weight and power (SWAP).86

FIG II-3: Sadowski Presentation

Mad Scientist Conference Day One

25
 China. China is also
Another player is China, long the leader in small consumer
investing heavily in advanced
drones. Chinese company DJI alone has around 70% of the
research,87 through both
global market, and now the Chinese military is seeing what they
overseas purchases as well as can do with this new technology. At an aerospace exhibition in
impressive domestic research December, state-owned China Electronics Technology Group
investments, particularly in the Corporation (CETC) displayed a video of nearly 70 drones flying
field of quantum computing a together. The drones flew in formation and collaborated in an
potential breakthrough enabler intelligence-gathering mission. Those drones could also
for artificial intelligence. China cooperate in a saturation attack on an enemy missile launcher.
has fielded armed UAVs and They all dive in to attack simultaneously from different
also developed unmanned directions far too many at once for the defenders to stop.
ground vehicles, such as the
By David Hambling, BBC (UK), April 27, 2017
Snow Leopard 10, which can
The Next Era of Drones will be Defined by Swarms (27 Apr 2017)
detect and detonate bombs.
According to the Defense
Science Board, "every major manufacturer for the Chinese military has a research
center devoted to unmanned systems.88

Chinese internet firm Baidu Inc has agreed to Violent Non-State Actors (VNSAs).
acquire U.S. computer vision firm xPerception for an Mad Scientists assessed that as these
undisclosed amount to support their renewed efforts technologies are increasingly embedded
in artificial intelligence xPerception, which makes into our human infrastructure, a wide
vision perception software and hardware with range of VNSA and super-empowered
applications in robotics and virtual reality, will individual threats become very feasible.89
continue to develop their core technology under
Terrorists are traditionally conservative
Baidu's research unit, the Chinese firm said in a
and imitative with respect to technology
statement on Thursday.
but today increasingly look to robotics,
"The acquisition of xPerception is the latest in a artificial intelligence, and autonomy for
recent series of notable investments aimed at multiple reasons.90 Some groups have
strengthening Baidu's position as a global leader in an ideological orientation towards
AI," it said. Baidu is targeting foreign personnel and technology either to leverage it or
technology as part of a wider drive to refocus undermine it. Many groups find existing
company resources on developing artificial methods insufficient to achieve their
intelligence capabilities. aims; these technologies lengthen the
Technology News, Reuters: 13 April 2017 levers of asymmetry. VNSAs confront a
need to circumvent protective measures
and make it more difficult to apprehend /
kill VNSA operatives. Use of these technologies reinforces the psychological impact of
terrorism, and also enhances the competitive status of the employer. As these
technologies permeate our infrastructure, a very high level of exploitable resources will
be available; e.g. driverless cars, and the costs associated with adopting new
technology are often low. The marginal cost to proliferate an AI capability through
software replication, for example, is close to zero.91

26
Some VNSAs are taking on complex engineering tasks and will train, hire or kidnap the
human capital they need to staff their R&D entities.92 Syria has emerged as an
innovation incubator that features VNSA use of drones, teleoperated rifles, remote gun
turrets and chemical weapons. ISIS recruits for engineers.93 Generally far less
burdened by acquisition bureaucracies, some adversaries are closing the technology
gap faster than we are widening it.94

Cyber and Dark Networks. Cyber networks are the Battlespace of Cyber Agents:
software code that incorporates automation and artificial intelligence to act in the cyber
domain.95 VNSAs frequently leverage Dark Networks, a component of the cyberspace
that is integral to the threat development of these technologies. Global Dark Networks
are mature, largely self-organizing and include competing supply chains that circumvent
regulatory controls.96

Algorithmic Warfare. Mad Scientists observed that conflict is extending below the
platform level, below the platform component level, and even below the electronic
chipset level as logic solutions compete algorithm vs algorithm. Algorithms are the
subtle secret sauce that powers these technologies, and underscores the need for
robust STEM programs to ensure the appropriate intellectual talent is available to
devise the most innovative, effective and efficient code. Deputy Defense Secretary
Bob Work has very recently established an Algorithmic Warfare Task Force to
accelerate the integration of big data and machine learning into DoD operations.97 98
Rogue Technology. Mad Scientists noted a high probability that the future will include
the threat of rogue technology that will be agile, high velocity, complex, networked, and
pop-up in unexpected, non-linear events.99 They further posited that future
adversaries may be both human and AI.100 Although hotly debated, there are a number
of pathways by which fully sentient artificial consciousness (strong AI) could be
achieved in the 2030-2050 timeframe.101 Early detection and control might be the only
available avenues to preclude existential rogue AI threats.

Technology Trends: Speed, Scope and Convergence

Consequential in their own right,
particularly in the hands of
adversaries, the impact of these
technology trends is exacerbated
by their collective speed, scope
and convergence.
Speed. Some Mad Scientists
posit that the rate of progress in
these technologies will be faster
than Moores law.102 As our
adversaries close the technology
gap and potentially overtake us in
The Need for Speed: the study concluded that autonomy will
deliver substantial operational value across an increasingly
diverse array of DoD missions, but the DoD must move more
rapidly to realize this value. Allies and adversaries alike also have
access to rapid technological advances occurring globally. In
short, speed mattersin two distinct dimensions. First,
autonomy can increase decision speed, enabling the U.S. to act
inside an adversarys operations cycle. Secondly, ongoing rapid
transition of autonomy into warfighting capabilities is vital if the
U.S. is to sustain military advantage.
DSB Report on Autonomy, 2016
27
select areas, there is clearly a need for speed as cited in the DSB Report on
Autonomy.103 The speed of actions and decisions will need to increase at a much
higher pace over time.104

Scope. It may be necessary to increase not only the pace but also the scope of these
decisions if these technologies generate the extreme future characterized by Mad
Scientist Dr. Ed Canton as hacking life / hacking matter / hacking the planet. In
short, no aspect of our current existence will remain untouched. 105 Robotics, artificial
intelligence, and autonomy far from narrow topics are closely linked to a broad
range of enabling / adjunct technologies identified by Mad Scientists to include:

o Computer Science, particularly algorithm design and software engineering

o Man-Machine Interface, to include Language / Speech and Vision
o Sensing Technologies
o Power and Energy
o Mobility and Manipulation
o Material Science to include revolutionary new materials106
o Quantum Science
o Communications
o 3D (Additive) Manufacturing
o Positioning, Navigation and Timing beyond GPS
o Cyber

Science and Technological Convergence. Although 90% of the technology

development will occur in the very fragmented, uncontrolled private sector,107 there is
still a need to view robotics, artificial intelligence and autonomy as a holistic, seamless
system. Technology convergence was a recurring theme among Mad Scientists. They
projected that we will alter our fundamental thinking about science because of the
exponential convergence of key technologies including 108

Nanoscience and nanotechnology

Biotechnology and Biomedicine
Information Technology
Cognitive Science and Neuroscience
Quantum Science

This convergence of technologies is already leading to revolutionary achievements with

respect to sensing, data acquisition and retrieval, and computer processing hardware.
These advances in turn enable machine learning to include reinforcement learning and
artificial intelligence. They also facilitate advances in hardware and materials, 3D
printing, robotics and autonomy, and open-sourced and reproducible computer code.
Exponential convergence will generate extremely complex futures that include
capability building blocks that afford strategic advantage to those who recognize and
leverage them.109

28
Co-evolution. Clearly humans and these technologies are destined to co-evolve.
Humans will be augmented in many ways: physically, via exoskeletons, perceptionally,
via direct sensor inputs, genetically, via AI-enabled gene-editing technologies such as
CRISPR, and cognitively via AI COGs and Cogni-ceuticals.110 Human reality will be
a blended one in which physical and digital environments, media and interactions are
woven together in a seamless integration of the virtual and the physical.111 112 As
daunting and worrisome as these technological developments might seem, there will
be an equally daunting challenge in the co-evolution between man and machine: the co-
evolution of trust.

Brave New World?

Billionaire entrepreneur Elon Musk on Thursday confirmed plans for his newest company, called
Neuralink Corp. a startup that aims to merge computers with brains so humans could one day engage
in consensual telepathy. Neuralink aims to implant tiny brain electrodes that first would be used to
fight brain conditions but later help humanity avoid subjugation at the hands of intelligent machines.

Mr. Musk has spoken out about the dangers of being left behind by the advancements of artificial
intelligence. The pace of progress in this direction matters a lot We dont want to develop digital
superintelligence too far before being able to do a merged brain-computer interface.

Mr. Musks comments come a day after Facebook Inc. announced similar ambitions. What if you could
type directly from your brain? asked Regina Dugan, who runs Facebooks secretive hardware division
Building 8, during a keynote address at the companys F8 developer conference Wednesday. Facebook
job postings show the company is looking to hire engineers to work on brain-computer interface
technology.
Rolfe Winkler, Wall Street Journal
20 April 2017

Trusted man-machine collaboration will require validation of system competence, a

process that will take our legacy test and verification procedures far beyond their current
limitations. Humans will expect autonomy to be nonetheless directable, and will
expect autonomous systems to be able to explain the logic for their behavior, regardless
of the complexity of the deep neural networks that motivate it. These technologies in
turn must be able to adapt to user abilities and preferences, and attain some level of
human awareness (e.g., cognitive, physiological, emotional state, situational knowledge,
intent recognition).113

29
This Page Intentionally Left Blank

30
III VISUALIZE: Solution Approaches that address the
Characteristics of the Future Operational Environment

Five potential solution approaches emerged as dominant themes throughout the

Conference as Mad Scientists discussed how the technologies of robotics, artificial
intelligence and autonomy might address the characteristics of the emerging Future
Operational Environment (FOE).

Visualize: Solution Approaches to the Emerging Operational Environment

Manned-Unmanned Teaming (MUM-T)
Asymmetric Awareness & Decision
Swarming
Intelligent Networks for the Internet of Battle Things
Autonomous Sustainment

Overview: the Characteristics of the Future Operational Environment

The characteristics of the emerging operational environment can be summarized as
follows: 114
Contested in all Domains. Competitors contest all domains, neither accepting nor
assuming sanctuary in any part of the land, air, sea, space, and cyberspace. Complex
and lethal engagements permeate the battlespace from the deep seabed to
geosynchronous orbit. Land, sea, air and space platforms encounter long range
precision munitions, highly accurate guided missiles, lasers and microwave weaponry,
stealthy and agile swarming robotic systems, and continuous probing of cyber systems.
This contest extends to both control and use of the entire electromagnetic spectrum.
Domain competition includes ongoing measures to degrade the effectiveness of AI,
battle management, and firmware targeted even down to chipset level in any platform or
weapon. This lethal exchange is not only characteristic within each domain, but also
between them: the range and precision of sensors and weapons allows routine cross-
domain engagement.
Unprecedented Speed. The momentum of human interaction unfolds at
unprecedented speed. The speed of many engagements laser systems, hypersonic
weapons, cyber-attacks -- far exceeds the reaction ability of normal humans.
Significant battle processes are highly automated and supervised by cognitively
augmented humans and man-machine centaur teams. Modern manufacturing
accelerates the rate of capability development so that by 2050, forces must have a very
dynamic capacity to adapt and integrate capabilities, both materially and cognitively.
WMD Proliferation. Military planning must account for nuclear weapons, fissile
materials, and chemical weapons, as well as novel and very dangerous biological
weapons derived from revolutionary advances in commercial biotechnology. WMD
proliferation is destabilizing as WMD haves coerce the have-nots. Several states
have not only secured WMD forces for rudimentary deterrent effect, but have achieved
31
credible and diverse retaliatory strike capabilities. Daunting in its own right, nuclear
proliferation complicates conventional operations: dual-purpose platforms and
command links pose serious escalation risk and complicate engagement decisions.
Complex Terrain the Norm. Urban environments sprawl horizontally, vertically and
socially, posing both challenges and opportunities. Land forces must operate in these
areas for sustained periods and now view such operations as the default expectation,
vice the exception. Some resort to the control by devastation urban techniques of
previous decades; others seek very precise, low collateral damage combat. Cities have
massive resources that can be directed for war, such as computer controlled machine
shops, 3D manufacturing facilities, small scale chip foundries, and a dense array of
consumer electronics, wireless nodes, and commercial and private fiber networks.
There is a premium on the ability to separate combatants from non-combatants in
dense urban environments; forces employ sophisticated human and cultural mapping,
biometric assessment and tagging at long range, and the ability to understand and
selectively control city services and utilities.
Hybrid Combatants. Competitors combine regular and irregular forces, paramilitaries,
terrorists, criminals and others to threaten strikes, raids, insurrection, information
operations or outright invasion when possible or advantageous. States integrate
manipulation of economic, financial, and political institutions to coerce, destabilize and
unbalance target states and societies around the world. These hybrid operations are
even more successful when they are augmented by conventional and WMD- escalation
strategies that deter or dissuade targets -- and their would-be partners.

Solution Approaches and the SciTech Crowdsourcing Exercise

During the conference, participants described multiple solution approaches to address these
future Operational Environment characteristics. The SciTech Crowdsourcing exercise further
augmented the future menu of possible solutions. Throughout the two week event, SciTech
Futures Crowdsourcing participants developed, refined, and prioritized nearly 140 distinct ideas
related to robotics, AI, and autonomy. These are cataloged at Appendix B: SciTech Crowd-
Sourcing Insights of this report, as well as a description of the methodology and outcomes of this
event.

Each of the ideas developed and prioritized during the SciTech Crowdsourcing exercise were
aligned to the five solution approaches in this section (MUM-T, Asymmetric Awareness and
Decision, Swarming, Intelligent Networks for the Internet Battle of Things, and Autonomous
Sustainment), as well as the year the idea could be expected to be achieved given the pace of
technological change or other constraints, and are found in the relevant sections below.

Manned Unmanned Teaming (MUM-T)

MUM-T is the synchronized employment of Soldiers, manned and unmanned air and
ground vehicles, robotics, and sensors to achieve enhanced situational understanding,
greater lethality, and improved survivability. The concept of MUM-T is to combine the

32
inherent strengths of manned and unmanned platforms to produce synergy and
overmatch with asymmetric advantages.115

On highly lethal battlefields contested in all domains, MUM-T redistributes risk away
from our most valuable and irreplaceable asset: our Soldiers. As WMD proliferation
increases the likelihood of dirty battlefield zones, MUM-T will enable access and
assessment throughout the battlespace. In complex terrain environments, sensors and
shooters linked through MUM-T decrease the likelihood and impact of surprise. MUM-T
will enable speed of movement, particularly for transportation assets in dangerous
environments such as high-threat road conditions or high-speed nap-of-the-earth (NOE)
flight. MUM-T has potential application against the complete range of hybrid
combatants.

FIG III-1: Fountain Presentation

Mad Scientist Conference Day One

MUM-T Crowdsourcing Observations. Solutions associated with the MUM-T

approach accounted for some 13% (19 of 139) of the total ideas developed during the
crowdsourcing exercise. Of these, four ideas were represented in the top 30 ideas
relevant to the Army in the future (marked with ** in the table). The top idea, Man vs.

33
Machine: Human, AI, and Robotic Employment Optimization was the second highest-
rated idea in the entire exercise. This idea envisions a DOD enterprise-wide
approaches to evaluating the relative merits of human labor vice machine labor within
the DoD. This idea promises a structured method to leverage the disruptive effect of
robotics, AI, and autonomy and apply MUM-T and machine labor in an optimal way. It
was imagined to be available to the force by exercise participants by 2020.

The full set of MUM-T crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Achieved
Manned Unmanned Teaming: Idea Name
2020 Man vs. Machine: Human, AI, and Robotic Employment Optimization**
2020 Sort Robotic Wearables: Autonomous Tourniquet
2020 Autonomous Mine Removal
2025 Mobile Protected Firepower
2025 Cyber Manifesto
2025 AWACS 3.0 Distributed Robotic Battle Management
2025 Appliqu Autonomy Kits
2025 AI-Assist for Combat Medic
2030 Walking Combat Vehicles
2030 Robotic Wingman**
2030 Combat Engineer Bots**
2030 Combat Robot Ethos
2035 Telecommuting to War**
2035 Second Skin
2035 Backup Brain
2040 Human Accessible Robot/AI Off Switch
2040 Remote Operated Assault Robots
2040 Enhanced Others
2045 Insect Man

Asymmetric Awareness and Decision (AA&D)

Asymmetric Awareness and Decision (AA&D) applies robotics, artificial intelligence,

and autonomous systems to regain and maintain situational awareness in complex
environments and overwhelm an adversarys command and control operational

FIG III-2: Fountain Presentation (Modified)

Mad Scientist Conference Day Two

34
tempo.116

AA&D can extend Situational Awareness across multiple contested domains, and to
lower echelons of C2 in the land force. AA&D is sorely needed to cope with the
unprecedented speed of unfolding events on future battlefields. Although prediction is
never possible, WMD proliferation and the increased potential of the use of CBRNE
weapons will put a premium on the best possible indicators and warning. 117 With
complex domains the norm, advanced visualization techniques will be necessary to
make sense of the large and rapid streams of data.118 In the face of hybrid combatants,
asymmetric awareness and decision must leverage complex and multi-dimensional
streams of input. Patterns of life, for example, are also in the cyber domain.119

Asymmetric Awareness and Decision Crowdsourcing Observations. Solutions

associated with the AA&D approach represented the single most popular category for
ideas during the exercise, accounting for nearly 24% (33 of 139) of the total set. Eight
of these ideas were represented in the top 30 ideas relevant to the Army in the future
(marked with ** in the table). The top idea, Machine Augmentation to Staff Functions
was the fourth highest-rated idea in the entire exercise. This idea described the
application of deep learning and AI to optimize thousands of courses of action, from
logistics, to personnel, intelligence, and operations fed from thousands of sensor and
ISR fsources. This idea would allow staffs to operate within adversary decision cycles
and allow the Army to seize and retain the initiative during operations. This idea was
imagined to be available to the force by exercise participants by 2025.

35
The full set of AA&D crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Achieved
Asymmetric Awareness and Decision: Idea Name
2020 Multi-Layer Multi-Spectral Lens Protection**
2020 Real News Aggregator
2020 A.I Assisted Searchable Portable Military Library Laptop
2020 Pocket Augmented Reality Real-Time Training
2025 Adversaries Simulating Us
2025 Autonomous Sensor Defeat
2025 Heads-Up Glasses, Dash, and Desk Displays
2025 Pocket Interactive Doctrine, Training, and Policies
2025 Anti-Autonomy Sensor Disruptors
2025 Military/Law Enforcement Rehearsals
2025 Kinetic Courier / Kinetic Jammer
2025 Multi-Mode Laser Designator
2025 Machine Augmentation to Staff Functions**
2025 Robotic Subterranean Operations
2025 AI Robotic Information Warriors
2025 Adaptive Hyperspectral Algorithm for Camouflage Detection
2025 Recon-by-Wire**
2030 Chatbot: AI Resurrected Clones of Great Thinkers
2030 EW Applied to Human Perception
2030 Cybernetic Super-Sniffers
2030 Misinformation Disintegrator
2030 Anti-Pattern Recognition Camo**
2030 Mesh Networks as Alternate Internet**
2030 Recon Round**
2030 Genetic Algorithms for Optimizing Team Composition
2030 Rent-an-Avatar Booth
2030 Counter-AI Operations Field Manual**
2035 Second Skin
2035 21st Century Non-Kinetic, Multidomain Training for All Troops
2035 TOC in a Box
2035 Ever-Present Commander Rules of Engagement Authority
2040 DigiPatton**
2045 Ultra-Fast Battlefield

36
Swarming Today, platforms rule the battlefield. In time, however,
the large, the complex, and the few will have to yield to
Swarms leverage autonomy, robotics the small and the many. Systems composed of millions of
and artificial intelligence to generate sensors, emitters, microbots and mini projectiles, will, in
concert, be able to detect, track, target, and land a
global behavior with local rules for
weapon on any military object large enough to carry a
multiple entities either human. The advantage of the small and the many will not
homogeneous or heterogeneous occur overnight everywhere; tipping points will occur at
teams. Collective behaviors emerge different times in various arenas. They will be visible only
because of simple rules at the in retrospect.
individual level that lead to complex M. Libicki (1996)
aggregate behavior. No individual in The Mesh and the Net:
the swarm knows the solution, but Speculations on Armed Conflicts
in an Age of Free Silicon
collectively, the swarm converges on
an optimal path or solution.120 Swarming leverages distributed awareness from multiple
sources; the multiplicity of the sources enhances learning.121 Swarm entities are
typically expendable and have limited broadcast and computing power; a swarm system
can survive loss of several platforms but still perform.122

FIG III-3: Autonomous UAS Swarming

Swarm tactics, techniques and procedures are feasible in all domains. Swarm
techniques will be one manifestation of the unprecedented speed of future operations
as multiple entities overwhelm linear, sequential decision processes.

Swarm Intelligence123 will search Big Data and seek to make sense of highly complex
environments. The difficulty of dealing with overwhelming attacks of the small, the
smart and the many may lead some competitors to leverage large scale area
suppression weapons, including weapons of mass destruction (WMD.) VNSA

37
components of hybrid adversaries frequently leverage swarm C2 techniques to
generate collective behavior without the use of interdictable / interruptible global C2.

Swarming Crowdsourcing Observations. Swarming-related ideas accounted for 8%

(12 of 139) of the total set. Two of these ideas were represented in the top 30 ideas
relevant to the Army in the future (marked with ** in the table). The top idea, Ten Cent
Defeat was the 14th highest-rated idea in the entire exercise. This idea described the
ability of all robots and autonomous systems to not fail spectacularly when confronted
with primitive, low-cost defeat mechanisms, adapt, and recover functionality. The idea
would apply a range of technologies and approaches to ensure that some percentage of
a robotic fleet would remain operational even when confronted with novel
countermeasures. This idea imagined to be available to the force by exercise
participants by 2020.

The full set of Swarming crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Swarming: Idea Name
Achieved
2020 Ten-Cent Defeat**
2025 Virtual Minefield
2025 Drone Swarms
2025 Mothership/UCAV Delivery Carrier
2030 Nano-AI Vaccines**
2035 AI Prototype Platform Design
2035 Autonomous Infrastructure Repair and Maintenance
2035 Swarming Attack Nano-Bots
2035 Permanent Protective Drone Swarms
2035 Nanobot/Microbot Tracing Sensors
2040 Sleeper Drones
2045 Attack of the Clones

Intelligent Networks for the Internet of Battle Things (IoBT)

Intelligent Networks are distributed, interconnected entities that have goals, sense
their environment, identify constraints and threats, and plan and execute adaptive
actions through the leverage of autonomy, artificial intelligence and robotics.124 The
Internet of Battle Things is not a straightforward extension of the Internet of Things,
because the Internet of Battle Things must deal with the adversarial nature of the
environment to include

Kinetic attack
38
 Directed Energy
Electronic Attacks
RF Channel Jamming
Destruction of fiber channels
Destruction or debilitation of power sources
Electronic Eavesdropping
Cyber Malware

FIG III-4: Kott Presentation

Mad Scientist Presentation Day Two

Intelligent Network applications in the emerging operational environment may include

networks of intelligent munitions passing data to defeat highly lethal adversary domain
defenses. High speed flight through highly cluttered, complex environments enabled by
dynamic, self-forming networks. We will see battles between networks of intelligent
munitions, and intelligent munitions might come sooner than intelligent platforms.125
Combatants will leverage the millions of devices embedded in megacities to sense and
deceive, and autonomous cyber agents in defense of mobile tactical networks will
leverage intelligent networks.

39
Intelligent Networks for the Internet Battle of Things Crowdsourcing
Observations. Intelligent Network-related ideas accounted for 13% (18 of 139) of the
total set. Three of these ideas were represented in the top 30 ideas relevant to the
Army in the future (marked with ** in the table). The top idea, Smart Dust was the third
highest-rated idea in the exercise. This idea described radio-frequency identification
(RFID) transmitters the size of a human hair with unique number strings for tracking
purposes which are deployed in varying amounts for discrete or mass surveillance. The
idea would provide a new range of ISR capabilities to the force to track and monitor
targets remotely and with high quality data. This idea imagined to be available to the
force by exercise participants by 2030.

The full set of Intelligent Networks crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Intelligent Networks for the Internet of Battle Things: Idea Name
Achieved
2020 AI-Enhanced Network Gate-keepers**
2020 AI Research Assistants
2020 Multi-Function Weapons
2020 Report Writer: Customizable AI Research Tool
2020 Robotic Programmers, Inc.
2025 Plug and Play Military Robotics Vehicles
2025 Corrupted R&D Simulation Software
2025 Kit to Control Captured Enemy Equipment
2025 Disrupter Bots for Crowd-Sourced Online Studies
2025 Algorithms to Approximate Human Judgments
2030 Teams of Small Robots to Move Casualties to Safety
2030 Anti-Machine Pathogens
2030 Smart Dust**
2030 AI Overrun Protection
2030 Neuronet
2030 Internet of (Hostile) Things
2030 Networked Autonomous Infrastructure Sabotage Battalion
2035 Machine Learning Pathologies**

Autonomous Sustainment

Autonomous Sustainment is a multi-modal approach that will leverage autonomy,

artificial intelligence and robotics to increase sustainment distribution, throughput and
efficiency.

40
Use of autonomous additive manufacturing can reduce forward sustainment demand
and minimize exposure to contested commons in the maritime and aerospace domains.
In a Future Operating Environment contested in all domains, autonomous (or leader-
follower) convoy delivery of supplies through areas of high threat, conventional or
hybrid. Autonomous Sustainment can leverage additive manufacturing and the
resources of highly complex environments such as large distributed urban areas to
minimize forward throughput and accommodate the unprecedented speed of operations
and associated logistic demands. Robotic delivery systems can also negotiate highly
contaminated areas generated by the use of Weapons of Mass Destruction.

FIG III-6
FIG III-5: Sadowski Presentation
Mad Scientist Presentation Day One

Autonomous Sustainment Crowdsourcing Observations. Autonomous

Sustainment-related ideas accounted for 8% (12 of 139) of the total set. Four of these
ideas were represented in the top 30 ideas relevant to the Army in the future (marked
with ** in the table). The top idea, Motorpool Bots was the 11th highest-rated idea in the
exercise. This idea described need to develop a capability to repair and maintain robots
in the future. Robots may significantly enhance PMCS as well as perform repairs and
system upgrades. Once they master the controlled environment, these systems could
then be outfitted with cross-country terrain mobility systems so they can follow units into
the field, repairing and recovering damaged system even under direct or indirect fire.

41
The idea would allow robots to undertake dirty, dull, and dangerous repair tasks for the
Army. This idea imagined to be available to the force by exercise participants by 2030.

The full set of Autonomous Sustainment crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Achieved
Autonomous Sustainment: Idea Name
2025 3D Printing for Maintenance Parts
2025 Additive Manufacturing Sustainment Brigades**
2025 Hoarder Drone
2025 Fabship Aircraft**
2025 Six Sigma Army Total Design and Maintenance
2025 Robotic CASEVAC**
2030 AI Based New Product Development
2030 Motorpool Bots**
2035 Walking Emergency or Construction Vehicles
2035 Integrated Electrical Logistics
2045 Autonomous Space Miners
2045 BN/BDE Experimentation and Upgrade Officer

42
IV DESCRIBE: the Competitions of Multi-Domain Warfare

Overview: the Competitions of Multi-Domain Warfare

The interaction of multiple great powers similarly equipped with emerging robotic, AI,
and autonomous technologies and simultaneously trying to address the strategic and
operational challenges of the future operational environment will drive a fundamental
change in the character of warfare, a change characterized as a series of
competitions.126

The simultaneous and interactive competitions of finders vs hiders and strikers vs

shielders will generate a battlespace of unprecedented lethality. Combatants will
struggle and aggressively innovate to generate survivable close engagement in the
face of formidable adversary range and lethality. The advantages of connection,
aggregation and centralization will trade against equally compelling motivations for
disconnection, disaggregation, and decentralization with the probable result of a
widely distributed, non-contiguous battlespace. In such a battlespace, at least between
peer competitors in the land domain, the defense will be relatively advantaged over the
offense.

Emerging capabilities in robotics, autonomy and artificial intelligence will present future
combat developers with interesting trade-offs between planning versus reaction, and
judgement versus autonomy. Competitors will have daunting escalation capabilities,
making escalation advantage a prominent feature of future force design, doctrine and
policy. The extended range and precision of land based systems will extend their
effects more routinely and more effectively into the other domains, so that legacy
combined arms synergy now extends across multiple domains. Similarly, there will be
widespread recognition that conflict is a competition, not only across every domain in
the physical dimension, but also across the cognitive and moral dimensions.

Describe: The Competitions of Multi-Domain Warfare

Finders vs Hiders
Strikers vs Shielders
Range & Lethality vs Close Engagement & Survivability
Disconnection / Disaggregation / Decentralization
vs Connection / Aggregation / Centralization
Offense vs Defense
Planning & Judgement vs Reaction & Autonomy
Escalation vs De-Escalation
Domain vs Domain
Dimension vs Dimension

43
Finders vs Hiders

In the emerging competition between finders and hiders, manned-unmanned teaming

(MUM-T) of Soldiers, platforms and sensors (finders) will extend the reach, ubiquity, and
safety of finders, extending their capabilities into complex and lethal environments.
Asymmetric Awareness and Decision (AA&D), enabled by AI-based signal processing
and big data techniques, will be applied to make sense of the plethora of sensor
information. On the other side of this competition, hiders will leverage inexpensive,
autonomous robots as decoys. Robotics, autonomy and artificial intelligence will enable
the dispersion of capabilities
across disaggregated entities
disrupting and concealing their
expected capability signatures.
Both sides of this competition will
leverage intelligent networks that
will autonomously maneuver to
minimize risk of intercept.
Intelligent, autonomous EW
assets will collect, assess and
react to new waveforms real
time. Swarming enabled by
autonomy and AI will serve as a
FIG IV-1: Sadowski Presentation
maneuver technique for both
Mad Scientist Conference Day One
hiders and finders. On highly
distributed and non-contiguous
battlespace, MUM-T can deploy sensors in depth, generate effects in depth, and
execute supporting actions that significantly expand a formations area of influence.127

Strikers vs Shielders

The finder-hider competition is fundamental because of the simultaneous maturation

and proliferation of the precision strike regime. The type of precision formerly
reserved for high end aero-space assets is extending to all domains and at every
echelon of engagement, including the individual Soldier. Combatants including many
non-state entities leverage multiple manifestations of precision strike: kinetic
weapons, hyper-kinetic weapons, directed energy, EMS, and cyber. Peer competitors
cannot long endure the application of such strike effects, particularly when directed by
robust find capabilities. Successful combatants devise shields: joint, combined arms
endeavors that target opposing finders, their linkages to strikers, or the strikers
themselves. For every manifestation of striker, there is a shielder counterpart: intercept
missiles, railguns, lasers, jamming. Robotics, AI and autonomy will play a key role in
the future competition between strikers and shielders.

44
The simultaneous and interactive competitions of finders vs hiders and strikers vs
shielders will generate a battlespace of unprecedented lethality. MUM-T will mitigate
the risk of movement through this lethal battlespace. Strikers will employ swarming
solutions to overwhelm and defeat shielders -- potentially at significant cost advantage.
Strike munitions connected via intelligent networks will collaborate to assess and defeat
sophisticated shield systems (See FIG IV-2).128 Shielders will employ autonomous AI to
generate Asymmetric Awareness & Decision (AA&D) and the speed of decision needed
to identify striker threats and allocate response resources. Autonomous Sustainment
will be needed to restore the inevitable losses associated with the striker vs shielder
competition.

FIG IV-2: Kott Presentation

Mad Scientist Conference Day Two

Range & Lethality vs Close Engagement & Survivability

Combatants in future striker-shielder competitions will leverage range and lethality to

penetrate and overwhelm a shield or, conversely shield against strikers. A
complementary approach is close engagement. Close engagement disintegrates
integrated defenses, causing concealed forces to unmask and uncover, exposing them
to the finder and striker dynamic. The challenge is the approach. Close engagement

45
forces need superior range and lethality in the close fight, but they also need
survivability: the protection (and mobility) that allows them to maneuver through denied
areas to close with and defeat the highly lethal assets securing the adversary shield.
Thus ensues the competition between range and lethality vs survivability and close
engagement:

MUM-T of
disaggregated armored
platforms allows land
forces to close with the
enemy across zones of
very high lethality at
reduced risk to Soldiers
(FIG IV-3). Intelligent
networks exacerbate
lethality by extending
the range of sensors
(finders) and the
precision of weapons
(strikers). Asymmetric
Awareness and
Decision (AA&D)
leverages AI to counter
enemy IADS (Integrated
FIG IV-3: Sadowski Presentation Air Defense Systems) by
Mad Scientist Conference Day One enabling boundary
effect air mobility: high
speed, very low nap of the earth flight. Cyber agents across Intelligent Networks extend
the range and impact of effects in the cyber domain, with significant cross-domain
impact that reduces the exposure to highly lethal effects in the physical dimension.
Robotics, AI and autonomy enable the flexible and high-production manufacturing
techniques that make possible the generation of new applications of mass via swarms
of relatively inexpensive autonomous weapons that can negotiate lethal zones to close
with and destroy their targets. These same capabilities enable Autonomous
Sustainment and the generation of mass needed to withstand high attrition conflicts.

Disconnection / Disaggregation / Decentralization vs Connection /

Aggregation / Centralization

In the future operating environment, consolidation of forces enhances their connectivity,

aggregation, and control. Such consolidation reinforces the strength of the shield and
ability to mass effects. On the other hand, such consolidation poses extreme risk and
there is a countervailing impetus to disconnect to the extent possible -- from global
sensing networks, disaggregate formations, and accept significantly decentralized

46
control. In this competition, robotics, artificial intelligence and autonomy may mitigate
the difficult trades between control and risk.

Disaggregation, for example, will demand flexible, scalable force structures that can
accommodate joint and combined capabilities decentralized to extremely low levels. As
formations disaggregate and disperse across non-contiguous battlespace, a very
diverse set of joint and combined arms capabilities will migrate to lower levels. Leaders
and staffs of super-enabled small units will need to leverage AI, robotics and autonomy
to generate the Asymmetric Awareness and Decision (AA&D) needed to generate joint
and combined arms synergy at increased levels of scale and complexity, but at lower
echelons of command and control. MUM-T may apply new manufacturing advances
and autonomy and afford command posts that are self-driving, fully connected server
farms and fully mobile.

The non-contiguous future

battlespace will pose two
particular challenges: the
contest to communicate and
the struggle to sustain.
Intelligent Networks will
adapt to attempts to interdict
communications.

Autonomous Sustainment
enables resupply on a
distributed, non-contiguous,
and very lethal battlefield.
The same advanced
manufacturing techniques
that enable Autonomous
Sustainment also enables
reduction demands by FIG IV-4: Kott Presentation
fabricating MTOE parts and Mad Scientist Conference Day Two
small arms munitions in
theater. They may also reduce evacuation requirements because of their ability to
engineer human organs and synthetic fluids. Advanced manufacturing techniques may
be able to produce tracked and wheeled vehicles with equivalent protection and
capabilities at weight reductions ranging from 50% to 90%, reducing logistic demands
for distributed, dispersed formations.129

Offense vs Defense

Offense vs Defense is a timeless competition, the outcome reflecting the strategic and
technical conditions peculiar to each era. In the emerging operational environment,
conditions favor the defense. With peer competitors robustly but equally -- equipped
with finders and strikers, the combatant who moves particularly over extended

47
strategic and operational distances -- is disadvantaged. A defensive stance favors the
development of more robust shields with robust passive sensors (finders), and offers
the advantage of hardened, redundant locations in the lethality vs survivability contest.

This inherent defense advantage will reward competitors who can generate strategic
surprise and present unprepared adversaries with a fait accompli. Asymmetric
Awareness and Decision (AA&D) may detect very subtle indicators and warning to aid
in the anticipation of preemptive attacks and the
prevention of such strategic surprise. Closing with
10 years from now if the first person
the enemy will be the tactical challenge of the era;
through a breach isnt a robot, shame on
us
MUM-T will limit losses as forces move across
highly lethal defensive zones. Swarms will be
Robert Work
employed to overwhelm and saturate defenses at
decisive points. Intelligent Networks will learn
real time and reconfigure during operations and even during engagements. Adversaries
will pursue forward presence in potential regions of conflict, particularly a forward
presence that support a prepared operational defense and its consequent advantages.
As these forward positions weather initial attacks, Autonomous Sustainment will serve
to extend their endurance until reinforced by follow on forces.

Planning & Judgment vs Reaction & Autonomy

The duel for initiative is inherent to the nature of war, but this duel has a unique
character in the emerging operational environment. Operational tools will work at
extraordinary speed and reach, and not infrequently precipitate unexpected
consequences. The planning paradox in the offense is that rapid execution depends on
very careful planning and condition setting, particularly in the cyber domain. On the
defense, however, faced with bewilderingly short reaction windows, many combatants
will resort to automated and increasingly autonomous decision processes.

Taking human beings out of loop poses potential advantages versus competitors
unwilling or unable to automate key decision processes. Automation will extend beyond
platforms to decision making itself.130 Asymmetric Awareness and Decision (AA&D) will
apply AI and decision aids to enable sub-millisecond decision for allocation of
information assets and direction of kinetic attacks across all domains.131 This
coordination will extend across many automated decision-making entities, most of it
without human knowledge or intervention.132 Distributed, autonomous cyber agents
working across intelligent networks will exercise a vast preponderance of the decision-
making for defensive cyber.

48
Escalation vs De-Escalation

Means of violence will be readily available to a wide range of actors in the future
operating environment, and on unprecedented scales. Many competitors will have
Weapons of Mass Destruction (WMD) and incorporate their routine use into doctrine
and policy. Conventional and cyber capabilities are so potent, moreover, that they can
generate effects on the scale of Weapons of Mass Destruction (WMD). The
competition between violence escalation and de-escalation will be central to stability,
deterrence, and strategic success. Robotics, AI and autonomy technologies will afford
numerous solution approaches that facilitate transition up and down the escalation / de-
escalation ladder.

MUM-T solutions will enable forces greater access to inhospitable CBRNE

environments. There will be a premium on pre-empting escalation to the employment of
WMD, so Asymmetric Awareness and Decision (AA&D) capabilities will enhance the
ability to capture early indicators and warning. Automated Sustainment to enable
sustained operations in such environments even when necessary to traverse highly
contaminated zones. Intelligent Networks reconfigure to ensure communications in
spite of large scale infrastructure disruption due to EMP or area kinetic effects
(thermobarics).

Domain vs Domain

The competitions of future warfare will extend and intensify in new domains, particularly
space and cyberspace. Each domain will be fiercely contested, and between great
powers assured dominance or even lasting advantage in any domain will prove elusive.
As the tools of warfare extend their physical capability to both find and strike, armies will
not constrain their planning and operations merely to the land domain. Each domains
unique physics will constrain platforms and techniques, but the highest art of combined
arms warfare will be to generate effects from one domain against another: leveraging
their relative advantages and mitigating their innate vulnerabilities. In this environment,
effective joint synergy will not be a bonus, it will be table stakes for survival.

Robotics and autonomous systems will extend the reach of legacy systems and extend
effects from one domain into another. Because of risks across multiple domains,
robotics, AI and autonomy are applied to disaggregate and decentralize key capabilities
(e.g., nuclear C2) while retaining control and function.133 The complexities of generating
multi-domain synergy across multiple domains impose daunting cognitive loading on
leaders to simultaneously accommodate the range of maritime functions, the speed of
air / space / cyber operations, and the tactical complexity of land warfare.134 Asymmetric
Awareness and Decision (AA&D) helps understand, visualize, describe and direct action
across multiple domains.

49
Dimension vs Dimension

Future conflict will be a competition, not only across every domain in the physical
dimension, but also the cognitive dimension, and even the moral dimension of belief
and values. Adversaries equally enabled by ubiquitous sensors, big data techniques,
responsive space satellites and robust social media access will enjoy competitive
levels of situational awareness. Information will be weaponized, directly through cyber
techniques or implicitly through social media techniques.135 There will be a recognized
premium for understanding and appreciation of the belief systems that motivate actors
in the moral dimension of conflict.

Robotics, artificial intelligence, and autonomy will enhance the significance of the
cognitive dimension of conflict, and extend the cognitive dimensions impact on the
physical dimension. The impact of these technologies, however, is so extensive that it
also tests the moral dimension of conflict: the dimension of our values and beliefs.
Between future competitors, a severe asymmetry of ethics may emerge with respect
to the limits of allowable human performance enhancement, permissible levels of
control for artificial intelligence and autonomous systems, delegation of authorities for
cyber attacks, the legitimacy of terror tactics, and a willingness to put noncombatants at
risk through WMD use -- or use of conventional means with equivalent impact.
Meanwhile, developments in robotics and artificial intelligence will render large groups of
armed humans less and less important in warfare. Already, the United States is turning to
robots and drones to accomplish tasks that just decades ago required humans. We use
robots to disarm bombs and check for threats inside buildings; we use drones to monitor
substantial swaths of territory, vacuum up electronic communications, and fire missiles at
designated targets. Already, military robots and drones are getting smaller, stealthier, hardier,
and smarter; within a decade or two (at most), the United States or some other state will
develop robots, drones, or other weapons systems capable of operating autonomously in
circumstances too challenging for humans to handle

perhaps we shouldnt be too quick to celebrate the end of war. Fewer piles of corpses is a
good thing, but wars diminishing brutality may also diminish political inhibitions, tempting
the powerful to employ their bloodless new mechanisms for coercion and control more
frequently and more indiscriminately, with fewer legal and political safeguards. The less
bloody wars of the future may spread more insidiously and invisibly, enabling ever-more
Orwellian forms of political control and because they barely resemble traditional wars, they
may be that much more difficult to discern and regulate.

Rosa Brooks Can There Be War Without Soldiers?

Mad Scientists observed that nothing in the current laws of war mandate human
decision-making vice machine decision-making. The laws of armed conflict cover
effects on the battlefield (e.g., proportionality, discrimination, precautions in attack). If a
machine can be used in a manner that meets these criteria, then it can be used lawfully.
However, one important asymmetry between people and machines under the laws of
war is that machines are not legal agents. Humans are bound by the laws of war; robots
/ autonomous systems are not combatants.136

50
V DIRECT: the Drivers of Outcome
As future competitors leverage the technologies of robotics, artificial intelligence, and
autonomy to wage the competitions that will characterize future warfare, success will
accrue to those most effective in the institutional contests that are already underway
and will shape the outcomes of the future. Those contests will include strategies and
policy, concepts, innovation and adaptation, combinations and learning.

Direct: The Drivers of Outcome

Strategy & Policy
Concepts
Innovation & Adaptation
Combinations
Learning

Strategies & Policy

Third Offset Strategy. DoD is already focused on the potential of these technologies.
In November 2014, thenSecretary of Defense Chuck Hagel announced a new Defense
Innovation Initiative, which included the Third Offset Strategy. The initiative sought to
maintain U.S. military superiority over capable adversaries through the development of
novel capabilities and concepts. Prior Secretary of Defense Ashton Carter built on
Hagels vision of the Third Offset Strategy, and Deputy Secretary of Defense Bob Work,
has championed numerous small bets on advanced capability research projects
featuring robotics, artificial intelligence, and autonomous technologies.137

US Army Robotics & Autonomous Systems

(RAS) Strategy. The US Army has also been
active in exploring the potential of these
technologies, and published a Robotics &
Autonomous Systems (RAS) Strategy in 2016. Dr
Robert Sadowski, Army Chief Roboticist and Chair
of the RDECOM Robotics Community of Practice
summarized this strategy in some detail for the
Mad Scientist Conference. The RAS Strategy
established five priorities as follows:

o Increased Situational Awareness

o Lighten Soldiers Physical and Cognitive
Workloads
o Sustain the Force with increased
distribution, throughput and efficiency
o Facilitate Movement and Maneuver
o Protect the Force

51
The RAS priorities are phased over time as shown at FIG V-1.

Near Term Priorities: Mid Term Priorities: Far Term Priorities:

Until 2020
ADAPT NEAR
Increase situational
awareness for
dismounted forces at
lower echelons
Improve sustainment
with automated ground
resupply
Facilitate movement with
improved route
clearance
Protect the Force with
EOD RAS platform and
payload improvements
2021-2030 Beyond 2030
EVOLVE MID INNOVATE FAR
Increase situational Increase situational
awareness advanced, awareness with persistent
smaller RAS and reconnaissance from
swarming swarming systems
Improve sustainment Improve sustainment with
with fully automated autonomous aerial cargo
convoy operations delivery
Improve maneuver with Facilitate maneuver with
unmanned combat advancements to
vehicles and advanced unmanned combat
payloads vehicles
FIG V-1: RAS Priorities by Phase138

The RAS Strategy Implementation Framework includes 3 Lines of Effort (LOE) as

described below at FIG V-2.139

LOE 1: Envision the Future

Validate potential formational constructs through soldier in the loop simulation via Early
Synthetic Prototyping (ESP) and User Wargaming
Explore different platforms/employment: purpose built/appliqu/optionally manned,
lethality/survivability/mobility mix, levels of supervision and degraded comms effects
Leverage Soldier Innovation Workshops to generate new concepts from Soldiers
Generate future CONOPS and requirements documents

LOE 2: Develop RAS Capability

In Theater Ground / Aerial Autonomous Resupply
Initial Virtual RAS Proving Ground
Initial Teaming Behaviors w/ Combat Formations
Soldier / Crew Deployed ISE
Open Autonomy Software Architecture

LOE 3: Sustain Integrated Campaign of Learning

Work with TRADOC Centers of Excellence to deliberately conduct operational experiments with
RAS platforms to embed the User Community in the technology development process.
Determine the utility of RAS platforms through relevant operational assessment to both drive
future CONOPS/TTPs/Requirements as well as feedback information to RAS technology
development (Gaps/Use Cases)
Leverage RAS ICDT management structure to layout battle rhythm of M&S through COE Battle
Labs coupled with hardware experimentation

FIG V-2

52
Introducing robotics technologies into the formation is fundamentally new for the Army
and requires concurrent technology development, operational experimentation and
CONOPS development in order to maximize the capability offered by autonomous
systems.140

Policy Ethics. The ethical issues underscoring policy decisions are not as
straightforward as simply Should we build a Terminator? There will be a plethora of
subtle operational issues to be confronted in system design and operation.141 Mad
Scientists explored several of these. If we have a loitering munition or an unmanned
vehicle that is operating in a communications-denied environment, how much autonomy
do we want to give it to attack emerging targets of opportunity? How much prior
information do we expect military commanders to have about specific targets for attack?
How much specificity about target selection can they delegate to a machine? If we
have an unmanned vehicle in a communications-denied environment and it is attacked,
do we want it to use force to defend itself? What about preemptively?142

The ethical issues associated with these technologies are global. Over 60 non-
governmental organizations -- part of a Campaign to Stop Killer Robots -- have called
for a legally-binding treaty banning autonomous weapons. Over 3,000 AI and robotics
experts signed a letter in 2015 calling for a ban on offensive autonomous weapons
beyond meaningful human control. For the past three years, nations have discussed
autonomous weapons in the United Nations Convention on Certain Conventional
Weapons; discussions may move to a more formal Group of Governmental Experts next
year, but there is currently no momentum towards a treaty. Only a handful of states (and
no major powers) have said they support a ban.143

Mad Scientists were generally One of the places that we spend a great deal of time is
in agreement that the key determining whether or not the tools we are developing absolve
decision is: what role do we humans of the decision to inflict violence on the enemy. And that
want humans to play in use-of- is a fairly bright line that were not willing to cross. it is entirely
force decisions? If we could possible that as we work our way through this process of bringing
automate everything, what enabling technologies into the Department, that we could get
decisions would we still want dangerously close to that line. And we owe it to ourselves and to
humans to retain, and why? 144 the people we serve to keep it a very bright line.
The Army is currently
GEN Paul Selva, VCJCS (Aug 2016?)
committed to systems that are
not totally autonomous:
Soldiers will always be involved to address lethal decisions.145 The future ethical
challenges for the Army and for the Nation will emerge when we contend, for vital
national interests, with peer competitors who have chosen to not symmetrically self-
constrain.

Concepts

An operational concept is an image of combat: a concise visualization that portrays the

operational challenges of adversaries and their capabilities, and the scenario by which

53
they will be defeated. Since the most effective operational concepts incorporate the
dominant socio-economic and technical trends of their time, we can expect future
concepts to feature solutions that leverage robotics, artificial intelligence and autonomy.
The Joint Community, in fact, developed the Joint Concept for Robotic and Autonomous
Systems (JCRAS) in 2016 to describe a joint vision of future robotic and autonomous
systems in use by 2035. Significantly, the JCRAS admits a unique focus on capability
development vice capability employment, and states a primary purpose to guide
comprehensive development across the Joint Force.146

The JCRAS presents a vision wherein, by 2035, the Joint Force will employ integrated
Human-RAS teams in a
wide variety of In May 1940, the Allied and German Armies squared off in what
was expected to be an extended campaign for the conquest of
combinations to expand France. Six weeks later, the victorious German Army marched down
the Joint Force the Champs-Elysees in Paris. How was it that the Germans, with
commanders options. fewer tanks, fewer trucks, fewer troops, less artillery and access to
This concept envisions a roughly equivalent technologies, managed to accomplish such a
future Joint Force that remarkable feat? While leadership, luck, and a host of other factors
capitalizes on were at play, the decisive factor was the remarkable way in which a
technological advances few German inter-war military thinkers envisioned and developed a
new way of warfare, known to the Allies as the blitzkrieg. German
to embed highly-capable
doctrine successfully integrated current technologies in aircraft,
and interconnected RAS radios, and tanks into a coherent and integrated way of fighting and
into every echelon and then applied it to great effect. The result was amplified because the
formation. It projects Germans fought an enemy that in many cases failed to account for
RAS evolution from tools the possibilities enabled by the new combination of these
for basic tasks into team technologies.
members capable of
coordinating and We are now on the cusp of a similar revolution in warfare with the
opportunity to integrate several current and near term technologies
collaborating across
into our concept of how we will conduct military operations in the
domains and Services. not-to-distant future. The winner of the next conflict will not likely
RAS will play a central be determined primarily by the state of their technologies, but by
role in performing and how well a nations military thinkers conceptualize future warfare in
supporting an extensive an integrated manner and then apply robotic systems, or warbots,
array of complex mission appropriately to our way of fighting.
sets across the range of Brian M. Michelson, The Coming
military operations. 147 Warbot Revolution (6 Mar 2017)

It is not clear that the Army will need to publish a developmental concept at its own
level; the JCRAS may suffice for that purpose. The representative solution
approaches presented earlier (MUM-T, Asymmetric Awareness and Decision,
Intelligent Networks, Swarming, and Autonomous Sustainment are likely candidates for
incorporation into broader concepts, doctrine, or tactics, techniques and procedures
(TTP).

Robotics, artificial intelligence and autonomy may also significantly reinforce legacy
concepts and doctrine. Mission Command, for example, is often cited as a
foundational concept for Centaur Teaming:

54
 the U.S. Armys concept of mission command gives it a decisive advantage on the
battlefield without even the expectation of perfect control over every soldier and vehicle.
While absolute autonomy is correctly viewed as unacceptable, if we are to see the full
benefits of warbots, we already have the mission command framework and can adapt it to
the new forms of interaction that will emerge between manned and unmanned systems.148

Innovation & Adaptation

Mad Scientists consistently stated that robotics, artificial intelligence and autonomy
technologies enable previously infeasible solution approaches and therefore may
change the character of warfare itself. Because these technologies will be broadly
available to most future combatants, the contest of innovation and adaptation between
those combatants will be a key driver of future outcomes.

Scope and Flow of Innovation. Mad Scientists asserted that innovation and
adaptation must comprehensively and simultaneously explore solution approaches from
the conceptual to the platform / sub-system level. Innovation and adaptation will not
only mitigate battlefield vulnerabilities and enable exciting new capabilities, it also offers
the potential for significant cost reductions in manufacturability and sustainment.
Robotics, autonomy and artificial Intelligence can improve the performance of legacy
platforms and will probably first be applied for incremental improvements to current
capabilities.149

Unlike for previous defense-related

technology waves, the flow of technology
particularly for Third Offset technologies
such as robotics, artificial intelligence and
autonomy -- is predominantly from the
commercial sector to defense. Therefore
most of these technologies have a dual-
use nature that makes them widely
accessible to both emerging great powers
and non-state entities.150 Think tank
studies such as the CNAS report Future
Foundry: A New Strategic Approach to
Military-Technical Advantage have
recommended optionality strategies in
which DoD evaluates a diverse array of
technologies in light of a greater volume of
high- and low-end threats. Focusing
initially on prototype design and limited
production, the government can delay full
production until threats become more
defined and capabilities verified.151
Future advanced manufacturing capabilities
might enable such surge production.
the velocity and volume of 21st Century
technological advancement in the commercial
sector from renewable energy sources to
nanotechnology, to the vast and profound
integration of communication and information
technology, or wildly disruptive processes like 3-D
manufacturing have rendered the Pentagon a
net importer of technology vs. the vaunted
exporter it had been for most of its existence. Left
unmanaged, the inability of the Pentagon to
quickly acquire and adapt a wider range of
technology will yield a dominant focus on
incremental improvements to existing systems at
best and will block military superiority at worst
The Pentagons modernization strategy relies
heavily on networked-enabled and autonomous
learning systems, cloud computing, robotics, and
the state-of-the-art in cybersecurity. Yet, there is
no identified path to acquire some of the worlds
most exciting commercial technology.
William Lynn and Sean O'Keefe,
C4ISRNET, 1 May 2017

55
Asserting that commercial technology can address 15 percent of the US defense market
requirements, the same study recommends contracted Commercial Systems Adapters
missioned to continuously search worldwide for commercial technology with
promising defense applications that might be adapted or integrated for military
purposes.152

Manufacturability. In an operational environment featuring peer technical competitors,

these technologies enable the manufacturability that may provide the decisive
competitive edge: the ability to build the things we need once we know we need them
very fast, efficiently and with high reliability.153 Manufacturability will transition from an
approach that must choose between customization or high volume, contend with
disconnected supply chain, produce single function products, and test early state
prototypes before full production. Future manufacturing will evolve to customization at
any volume, a fully integrated supply chain, multi-functional products, and production at
any scale.154 One of the most important applications of the machine learning dimension
of artificial intelligence will be its application to the Industry 4.0s Manufacturing
Revolution.155 Machines that can Learn can inform Machines that can Make,
enabling solution approaches such as remanufacturing weapons based on the outcome
of the previous engagement.156

The Innovation Race. With widespread recognition that the race to innovate and adapt
is a contest with many competitors ranging from emerging great powers to super-
empowered individuals, there was a widespread recognition among Mad Scientists of
The Need for Speed. Echoing the words of Mad Scientist Dr James Canton, Dr
Alexander Kott stated:

Much is possible today so what are we waiting

for? If we wait, our adversaries will not wait and
the future may be coming towards us when we
least expect it. The future will come at us much
faster than we want. Better to disrupt ourselves
than be disrupted by adversaries.157

From a military perspective, innovation without

acquisition will not be relevant, and our legacy
acquisition systems are routinely identified as
FIG V-3: Kott Presentation
strategic handicaps. There are widespread
Mad Scientist Conference Day 2
recommendations to move away from high
quality acquisition strategies that frequently fail
after decades of investment, to a good enough approach that aims to fail early and fix
early. $100 million field tests where failure is not an option are problematic.158 Good
enough acquisition strategies could readily leverage machine learning, robotics and
autonomy for iterative yet efficient prototyping and production.

56
Mad Scientists noted that
developmental approaches for this
technology are predominantly
open source. Due to the vast
amount of available open research
and the rate at which it is
published, progress in
development and fielding of these
technologies is increasingly a
numbers game, particularly in the
fields of machine learning and
artificial intelligence. Competitors
need lots of science, technology,
engineering and mathematics
(STEM) skills to explore the huge
space of low-hanging fruit. 159 The
innovation race is also a STEM
race, but the gap between the
United States and its competitors with respect to STEM skills continues to close.
The United States still offers the worlds largest supply of
scientists and engineers, but countries in East and Southeast
Asia most notably China have been catching up, the
National Science Board said in its annual Science and
Engineering Indicators report, made public Thursday
afternoon.
Americas lead is distinct but decreasing, the board said.
China, it pointed out, almost tripled its number [of
researchers and science and engineering workers] since the
mid-1990s.
Meanwhile, from 2003 to 2012, Chinas high-tech
manufacturing sector grew fivefold, an increase that tripled
its contributions to global high-tech manufacturing from 8
percent of the market to 24 percent in just nine years.
US News and World Report. Feb 6, 2014

Innovation and adaptation is enabled by a set of developmental standards and

frameworks that are widely adapted across DoD and the Army and serve as
developmental speed enablers. Mad Scientists presented several, including:

Autonomous Ground Vehicle Reference Architecture Software Standard.

The AGVRA software architecture is constructed to enable a modular approach
to upgrade and acquire autonomous system behaviors for military ground
systems. The AGVRA bases military autonomous software on the world largest
open-source framework and development community to maximize opportunities
for innovation across industry, academia and the government.160 Subsequent
spirals will increase the level of modularity of the software, adding more
interfaces and enabling greater competition at lower behavioral levels. 161
Robotic Operating System (ROS) is a collection of software frameworks for
robot software development providing operating system-like functionality on a
node-cluster format. ROS is open source and used by the majority of robotics
developers in industry and academia. The Army is adapting ROS to be the
baseline software development framework for future autonomous behaviors by
creating ROS-Military (ROS-M) for military specific applications.
Robotic Technology Kernel Library. The RTK Library is a government
developed, ROS-based autonomy software application library that creates a
reusable foundation of autonomous platform behaviors that can be applied
across multiple mission roles and systems.162

57
 Unmanned Ground Vehicle Interoperability Profile (IoP). PEO CS&CSS
funds TARDEC to maintain and develop their IOP which defines software
messaging & hardware interfaces between major subsystems of unmanned
ground systems utilizing existing standards.163
These developmental, open source standards are both speed enablers but also
potential points of vulnerability that
illuminate design and control
architectures to potential fast
followers and hackers. Internal
encryption will be an important
requirement for the security of future
systems.

As we enter the early stages of a global

innovation economy, the biggest global
risk factor is not innovating fast
enough.164 In the words of Mad
Scientist Dr. James Canton, there are
new rules: Disrupt yourself before you
get disrupted.165
FIG V-4: Canton Presentation
Mad Scientist Conference Day 2
Combinations

Warfare has always been the art of combinations; throughout history combinations are a
significant driver of conflict because outcomes belong to the competitor most
imaginative and effective in presenting his adversary those combinations that pose
complex, multiple dilemmas. At the national level this entails effective integration of all
the elements of power: diplomatic, informational, military and economic. At the
operational level the art of combining diverse capabilities in multiple domains generates
joint and combined arms synergy. This set of technologies robotics, artificial
intelligence, and autonomy is a classic example of the power of combinations. They
are not only routinely applied interdependently; they have vastly enhanced potential in
combination with each other as well as with legacy or emerging technologies.

Mad Scientists explored both the potential and challenges of these combinations.
There is a particularly challenging combination challenge in the interaction between
man and machine. Man-machine combinations that are particularly challenging include
166

a human sensing using sensors of autonomous system, while controlling it via

augmented reality (remotely);
a human controlling a swarm or group of robots;
a human overseeing autonomous control of entire fleets;
a human specifying high-level objectives and automated agents enacting
those objectives.

58
Solving such challenges will entail significant conceptual and technical advances and
may still leave us with some persistent dilemmas:

Inferior Interfaces. Currently the interfaces between humans and machines

are relatively immature and not optimized for either component of the man-
machine team. Mad Scientists pointed to requirements for progress in natural
language processing and explainable AI: explanation of AI decision.167 DARPA
has initiated a program (Explainable Artificial Intelligence) to address this
need.168 169

Infinite Situations. There will be a bewildering and virtually limitless set of

complex and uncertain environments that can not possibly all be accounted for in
design. Man-machine combinations will need strategies to effectively and very
quickly transfer the appropriate learning to address new situations.170

System Integration. The solutions for sensing, perceiving, detecting,

identifying, classifying, planning for, deciding, and responding to a very diverse
and dynamic set of threats in a very complex environment will be innately
complex. The integration challenges will be daunting.171

Human Unpredictability. The human component of the man-machine team is

innately unpredictable and increasingly, so is the artificial intelligence
component.172

AI Unexplainability. Paradoxically, the application of these increased technical

capabilities may actually decrease our ability to predict and control outcomes.173

Learning

A Campaign of Learning for Robotics, Artificial Intelligence and Autonomy.

Because of the broad range of enabling technologies and sciences for robotics, artificial
intelligence and autonomy, Mad Scientists recommended a comprehensive campaign of
learning across a wide range of topics as shown below in FIG V-5.174 175 176

Manned Asymmetric Awareness & Machine Learning

Unmanned Teaming Decision and Artificial Swarming
Intelligence
o Unmanned o The Big-Data Storage o Unsupervised o Sensing and
Ground Vehicles and Architectures learning, Perception
o Unmanned Air o Analytic Algorithms generative o Navigation
Vehicles o Augmented Reality modeling o Communications
o Robust Situational Awareness o Reinforcement Architectures
Communications and Targeting learning for o Autonomous
o Reliable/assured o Small Unit Leader decision-making Behaviors
PNT Precision Targeting o Multi-task o Human Factors
o Man-machine o Integrated Sensor networks,
interface Architectures transfer learning

59
 o Lethality o 3D Enriched Urban o Multi-source o Trust and
o Platform Terrain Visualization fusion and Dynamic Team
Protection o Advanced Training and distributed Formation
o Manned Vehicles Simulation Technologies sensing o Heterogeneous
o Cyber resiliency o Wearable Devices o Human Teams
o Soldier Enhancement and performance
Optimization
FIG V-5 Technology Enablers for the Campaign of Learning

In addition to the broad array of enabling technologies in FIG V-5, Mad Scientists
asserted that the Campaign of Learning would be well served by basic research in
areas to include:

o Life Sciences
o Material science
o Nanotechnology
o Biotechnology
o Information Science
o Quantum Science
o Socio-cultural Behavioral Sciences
o Human Performance Augmentation

Learning to Learn. Researchers are

discovering techniques that enable AI software to
learn how to write AI software. Learning to
learn developmental approaches may
significantly reduce the typically huge volumes of
data needed to learn a specific task by
inference from solutions already developed for
similar but distinct problems. Such
techniques may accelerate the development and
fielding of the AI capabilities that enable our most
advanced robotic and autonomous systems.177
Google Brains researchers describe
using 800 high-powered graphics
processors to power software that came
up with designs for image recognition
systems rivaling the best designed by
humans
Tom Simonite AI Software Learns to
Make AI Software MIT Tech Review

Experimentation. In a future characterized by technological peer competitors,

experimentation will be increasingly important. Mad Scientist Dr Sadowski offered that:

Successful implementation of RAS will not predominantly be a hardware race whose

robot shoots farther or better to make these platforms truly members of the combined
arms team will require user experimentation to refine/guide S&T developmental paths and
enable the user to employ innovative CONOPS 178

Observing that artificial intelligence frequently succeeds when applied to a specific

space of objects and behaviors; Mad Scientist Dr Kott recommended more
experimentation, especially early in the development process. He advised that initial
objectives can be simple: Dont make the tech do everything the prior capability can.
Try single things, later, combine things if you can / must.179

60
Crowd-sourcing. Institutions in all domains, including the U.S. Government, are
increasingly turning to crowd-sourcing as a learning technique, and this Mad Scientist
event was notable for its inclusion of this technique in the SciTech Crowd-sourcing
Exercise.180 This crowd-sourcing exercise was essentially a game, and Mad Scientists
noted that ...

The superior human strategies stem from the minds ability to capture the essence of a
problem. Quantum concepts may seem less bizarre to people in a game than they do in
other contexts, because it is an environment in which they expect rules to be broken. 181

There is a notion that expertise is narrow but we see the power of crowd sourcing. When you
create open doors for people to showcase their work and introduce it into official channels it is
powerful and enduring.182

Results of the SciTech Futures Crowd-sourcing exercise are described at Appendix B.

The Art of Learning. Mad Scientist August Cole argued compellingly that artists
particularly fiction writers -- are a strategically underutilized asset for future foresight.183
The artistic creation of an idea often precedes the reality of novel ways of fighting.184
The ability to think creatively and holistically is critical: the most catastrophic failures are
the failures of imagination. We need to reach the young, we need to reach out
internationally, and we need to reach out to those not usually heard. Cole advocated
FICINT: an intelligence process that monitors fictional writing. He asserted that fiction
will be increasingly important for combat developers because the pace of events will
otherwise outpace us. How do you get to deep thinking? Creating the science fiction
world allows you to do that.185

TRADOC has several calls for papers, including a science fiction Warfare in 2035 and
Visualizing Multidomain Battle to assist in this deep thinking and provide current-future
contrast necessary for effective future learning.

The following extract from the Atlantic Councils Art of the Future illustrates the power
of fiction for imagining and envisioning the future.

The Death of Homer

Captain Stacy Doss felt the urge to yell to Homer, to ask him if he was OK. But
that was stupid, because he was dead, and that made it highly unlikely that he
would answer. Homer wasnt his real name, just a funny nickname for her favorite
platoon leader. She had lost contact with him a few hours ago. He was brave,
resourceful, and probably killed by an enemy warbot. And here she was, a Captain
in the United States Army, sitting north of Manila, feeling the sense of guilt only
commanders know after a battle. Call it the weight of command. Maybe she
missed something, perhaps she could have done better. As the commander of
Alpha Company, 1st Battalion, 15th Infantry Regiment, she needed to find him to
pay her respects, for closure. It was what commanders did

61
 Stacy finally found him, or what was left of him. He was the first one she
had seen when she arrived in the unit two weeks ago. Being young and new to
war, she had thought it would be cute to write Homer on the side of his hull to
personalize him a bit. It seemed like a shame that the supply system knew Homer,
and those like him, simply as an M316 Heavy Offensive Multi-Role Robot, or
HOMRR for short.
She shed no tears, but had a very strange sense of guilt that she was the
one who had sent him into a fight that he did not survive. He had fought bravely,
courageously, and autonomously. The correct term was no longer KIA, or killed
in action, but rather DIA, or Destroyed in Action. She had to remind herself that
Homer was a robot and that had he and those like him been manned fighting
vehicles, many more Soldiers would have died not only in this battle, but also in
this war.186

(Image by Alex Brady: Laser Tank @ http://artoffuturewarfare.org/2017/02/warbot-1-0-the-death-of-

homer/)

62
VI Conclusion
There is indeed the potential that developments in robotics, artificial intelligence and
autonomy will precipitate the Cambrian Explosion described at this reports
introduction. Efforts such as this Mad Scientist project are essential to appreciate the
scope, depth, and impacts of these transformative capabilities and the innovative
approaches they will enable. To the extent that we can advance that appreciation, we
mitigate the risk that this Cambrian Explosion becomes our Cambrian Conundrum.

To that end this project has sought to understand the fundamental trends propelling
these technologies from both a friendly and threat perspective, to include an
appreciation of their speed, scope, and convergence. Building on that understanding
with the insight of the Mad Scientists, we can visualize five principal solution
approaches that leverage these technologies to address the emerging characteristics of
the Future Operating Environment. We can further describe how solution approaches
such as MUM-T, Asymmetric Awareness and Decision, Swarming, Intelligent Networks
and Autonomous Sustainment can be applied to the competitions that will characterize
future warfare. The outcome of those future competitions will be the consequence of
how we currently direct the drivers of outcome: strategy and policy, concepts, innovation
and adaptation, combinations, and learning.

Mad Scientists noted the wisdom generated by predecessors who also peered into an
ambiguous future. Mad Scientist Dr. Gary Ackerman cited Thomas C. Schelling:

The danger is not that we shall read the signals and indicators with too little skill; the
danger is in a poverty of expectations a routine obsession with a few dangers that may
be familiar rather than likely.... The planner should think in subtler and more variegated
terms and allow for a wider range of contingencies. 187

The challenge, as our CSA has reminded us, is to be not too wrong. Along those lines
Mad Scientist Dr. Augustus Fountain cited Wick Murrays Adaptation in War:188

Those military organizations that display imagination and a willingness to think through
the changes that occur in the tactical, operational, and strategic levels in peacetime have
in nearly every case been those that have shown a willingness and ability to adapt and
alter their prewar assumptions and preparations to reality.189

In his closing remarks to the Mad Scientists, TRADOC G2 Tom Greco cited Disrupt
yourself before you are disrupted as the quote of the day. 190 Both Schelling and
Murray would approve. So would Dr Roper, Director of the Strategic Capabilities Office:

Our challenges and opportunities are great. Our challenges

because they require a new playbook, but our opportunities because
creating [a new playbook] is leveraging some of our nation's greatest
strengths -- ingenuity, technology and ... our unparalleled operators.
I like our chances.191

63
End Notes
1
Mad Scientist Conference 2016: Strategic Security Environment in 2025 and Beyond (October 2016), pp 12-13.
2
Build the Workforce is LOE 1, Partnerships is LOE 5 in GEN Mark Milley; Chief of Staff, U.S. Army The Army
Cyberspace Strategy for Unified Land Operations, January 2016, pp. i
3
The Cambrian Explosion in biology was a rapid diversification of animal body shapes and forms from simple,
immobile creatures to more complex, mobile creatures able to interact with their environment and each other.
See Gill Pratt, Is a Cambrian Explosion Coming for Robotics? IIEE Spectrum (31 August 2015).
4
The Burgess Shale Website @ http://burgess-shale.rom.on.ca/en/science/origin/04-cambrian-
explosion.php#unique
5
The International Weekly Journal of Science, nature website @ http://www.nature.com/news/what-sparked-
the-cambrian-explosion-1.19379
6
Andrew Ilachinski, Center for Naval Analysis, AI, Robots, and Swarms: Issues, Questions, and Recommended
Studies, January 2017.
7
Dr. Augustus Fountain, Deputy Chief Scientist (ST) Office of the Deputy Assistant Secretary of the Army (Research
& Technology), Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy:
Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
8
Mr Tom Greco, Opening Comments to the Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
9
Draft TRADOC Paper: The Operational Environment, 2035-2050: The Emerging Character of Warfare, March
2017.
10
The Joint Staff, Joint Concept for Robotics and Autonomous Systems (JCRAS), 19 October 2016.
11
Paul Scharre, Senior Fellow and Director of the Future Warfare Initiative at the Center for New American
Security, Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning
Multi-Domain Warfare in 2030-2050, 8 March 2017.
12
Sadowski, op cit.
13
Scharre, op cit.
14
Scharre, op cit.
15
Dr. Charles Pippin, Senior Research Scientist in the Aerospace, Transportation and Advanced Systems (ATAS)
Laboratory at Georgia Tech Research Institute, Presentation to Mad Scientist Conference: Robotics, Artificial
Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
16
Brianwieck in TRADOC G2 Mad Scientist SciTech Crowd-Sourcing Exercise:
https://scitechfutures.com/ex6/workshop/ai-robotic-information-warriors/
17
Scharre, op cit.
18
Brian Michelson, Brian M. Michelson, The Bridge: Blitzkrieg Redux: The Coming Warbot Revolution (6 Mar 2017)
19
Scharre, op cit.
20
Dr. Nahid Sidki, Executive Director of Robotics R&D, SRI International, Presentation to Mad Scientist Conference:
Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 8 March 2017.
21
Sidki
22
Sidki
23
Sidki
24
Laura Stevens and Tim Higgins, Wall Street Journal, Amazon Studies Driverless Ideas, Wall Street Journal
(Technology) p B4, 25 April 2017.
25
Dr. Jaime Carbonell, University Professor and Allan Newell Professor of Computer Science at Carnegie Mellon,
Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-
Domain Warfare in 2030-2050, 7 March 2017.
26
Sidki, op cit.
27
Sidki, op cit.
28
Sidki, op cit.
29
Sidki, op cit.
30
Wiley OnLine Library. Complexity, TightCoupling and Reliability: Connecting Normal Accidents Theory and High
Reliability Theory @ http://onlinelibrary.wiley.com/doi/10.1111/1468-5973.00033/abstract

64
31
Scharre, op cit.
32
Richard Potember, Perspectives on Research in Artificial Intelligence and Artificial General Intelligence Relevant
to DoD, JASON, The MITRE Corporation. January 2017.
33
Carbonell, op cit.
34
Dr. Zsolt Kira, Branch Chief of Advanced Machine Learning Analytics Group within the Robotics and Autonomous
Systems Division at the Georgia Tech Research Institute, Presentation to Mad Scientist Conference: Robotics,
Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
35
Juliane Gallina, Director, Cognitive Solutions for National Security (North America) IBM WATSON, Presentation to
Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in
2030-2050, 7 March 2017.
36
Dr. Alexander Kott, Chief, Network Science Division, Computational and Information Sciences Directorate, US
Army Research Laboratory, Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 8 March 2017.
37
Carbonell, op cit.
38
Scharre, op cit.
39
Scharre, op cit. citing a statement by Kevin Kelly in Wired Magazine, 2014: The Three Breakthroughs That Have
Finally Unleashed AI on the World, Wired Magazine, 27 October 2014.
40
Dr. James Canton, CEO and Chairman of the Institute for Global Futures, Presentation to Mad Scientist
Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 8
March 2017.
40
NPR Planet Money Podcast: BOTUS, Episode 763, April 7, 2017.
41
Nanette Byrnes, Goldman Sachs Embraces Automation, Leaving Many Behind, MIT Technology Review VOL
120 No 3, May /June 2017.
42
Kira, op cit.
43
Carbonell, op cit.
44
Louis Mazziotta, Armament Research, Development, and Engineering Center (ARDEC), Presentation to Mad
Scientist Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-
2050, 8 March 2017.
45
Scharre, op cit.
46
Scharre, op cit.
47
Carbonell, op cit.
48
Carbonell, op cit
49
One Mad Scientist Dr Jaime Carbonell advocates a major extension to active learning by incorporating a more
realistic inclusion of multiple, potentially-fallible or reluctant external information sources with variable costs and
unknown reliability. Proactive learning reaches out to these sources and jointly optimizes learning source
properties (e.g. labeler accuracy, expertise area), selection of source, and selection of maximally-informative
instances for the learning task at hand. The proactive sampling methods trade off cost vs. information value and
amortized benefit vs. immediate rewards, being largely agnostic to the base-level learning algorithms.
https://cs.byu.edu/colloquium/active-and-proactive-machine-learning
50
Scharre, op cit.
51
Scharre, op cit.
52
Dr. Gary Ackerman, University of Maryland, National Consortium for the Study of Terrorism and Responses to
Terrorism, Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning
Multi-Domain Warfare in 2030-2050, 8 March 2017.
53
Will Knight, The Dark Secret at the Heart of AI, MIT Technology Review VOL 120 No 3, May /June 2017.
54
Carbonell, op cit.
55
Kira, op cit.
56
Gallina, op cit.
57
Louis Mazziotta, op cit.
58
Carbonell, op cit.
59
Wilson Brissett, Not Enough People to Solve the Cyber Threat, Air Force Magazine Daily Report, 4 May 2017
60
The Joint Staff, Joint Concept for Robotics and Autonomous Systems (JCRAS), 19 October 2016.
61
The Joint Staff, Joint Concept for Robotics and Autonomous Systems (JCRAS), 19 October 2016.
65
62
Sidki, op cit.
63
U.S. Army, Robotics and Autonomous Systems (RAS) Strategy, Maneuver, Aviation, and Soldier Division Army
Capabilities Integration Center, U.S. Army Training and Doctrine Command, January 2017.
64
Derived from RAS, ibid.
65
Derived from RAS, ibid.
66
RAS, ibid.
67
Scharre, op cit.
68
Dr. Magnus Egerstedt, Executive Director for the Institute for Robotics and Intelligent Machines at the Georgia
Institute of Technology, Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy:
Envisioning Multi-Domain Warfare in 2030-2050, 8 March 2017.
69
Scharre, op cit.
70
Sidki, op cit.
71
Sidki, op cit.
72
Sidki, op cit.
73
Sidki, op cit.
74
Canton, op cit.
75
Egerstedt, op cit.
76
Egerstedt, op cit.
77
David Rotman, Desktop Metal Thinks Its Machines Will Give Designers and Manufacturers a Practicable and
Affordable Way to Print Metal Parts, MIT Technology Review VOL 120 No 3, May /June 2017.
78
Egerstedt, op cit.
79
Scharre, op cit.
80
Scharre on Centaur Warfighting (Temple University)
81
Bret Swanson and Michael Mandel, Robots Will Save the Economy, The Wall Street Journal, 15 May.
82
Mark Muro, Andrew McAffee and David Rotman in Letters to the Editor, MIT Technology Review VOL 120 No
3, May /June 2017.
82
Egerstedt, op cit.
83
Sidki, op cit.
84
Sidki, op cit.
85
Elena V. Vaganova, Department of Innovative Technologies, National Research Tomsk State University,
Indicators of Innovation Potential of the Country as Means of the Government Policy Modeling in the Dominant
and Emerging Technological Regimes. https://www.triplehelixassociation.org/wp-
content/uploads/2015/09/VAGANOVA-E.-Paper-1-TRIPLE-HELIX-2015.pdf
86
Dr. Robert Sadowski, Robotics Senior Research Scientist Research, Technology and Integration Director at U.S.
Army TARDEC, U.S. Army Chief Roboticist, Presentation to Mad Scientist Conference: Robotics, Artificial
Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
87
Cate Caddell, China's Baidu buys U.S. computer vision startup amid AI push Reuters Technology News, 13 Apr
2017 @ http://www.reuters.com/article/us-baidu-m-a-idUSKBN17F0JF
88
Michael Horowitz, Foreign Policy Magazine, The Looming Robotics Gap, 5 May 2014,
http://foreignpolicy.com/2014/05/05/the-looming-robotics-gap/
89
Ackerman, op cit..
90
Ackerman, op cit.
91
Ackerman, op cit.
92
Ackerman, op cit.
93
Ackerman, op cit.
94
Canton
95
Kott, op cit.
96
Canton, op cit.
97
Cheryl Pellerin, DoD News, Defense Media Activity, Defense Innovation Maintains Military Overmatch Against
Adversaries, 03 May 2017 @ https://www.defense.gov/News/Article/Article/1172099/defense-innovation-
maintains-military-overmatch-against-adversaries/
98
Marcus Weisgerber, The Pentagons New Algorithmic Warfare Cell Gets Its First Mission: Hunt ISIS, Defense
One, 14 May 2017.
66
99
Canton, op cit.
100
Ackerman, op cit.
101
Ackerman, op cit.
102
Canton, op cit.
103
Defense Science Board, Summer Study on Autonomy, June 2016.
104
Fountain, op cit.
105
Canton, op cit.
106
APMAS 2011 Conference Website, Revolutionary New Materials! @
http://www.apmas2011.org/revolutionary-new-materials.html
107
Canton, op cit.
108
Canton, op cit.
109
Canton, op cit.
110
Canton, op cit.
111
Canton, op cit.
112
Institute of the Future Web Site: Blended Reality: Superstructing Reality, Superstructing Selves @
http://www.iftf.org/our-work/people-technology/technology-horizons/blended-reality//
113
Sidki, op cit.
114
Extracted from The Operational Environment, 2035-2050: The Emerging Character of Warfare
115
Sadowski, op cit.
116
Modified from Dr. Augustus Fountain, Deputy Chief Scientist (ST) Office of the Deputy Assistant Secretary of the
Army (Research & Technology), Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
117
Kira, op cit.
118
Kira, op cit.
119
Sadowski, op cit.
120
Scharre, Paul. Senior Fellow and Director of the Future Warfare Initiative at the Center for New American
Security. Robotics on the Battlefield Part II: The Coming Swarm CNAS website @
https://www.cnas.org/publications/reports/robotics-on-the-battlefield-part-ii-the-coming-swarm
121
Kott, op cit.
122
Pippin, op cit.
123
Andrew Ilachinski, op cit.
124
From Kott, op cit.
125
Kott, op cit
126
Extracted from The Operational Environment, 2035-2050: The Emerging Character of Warfare
127
Sadowski, op cit.
128
Kott, op cit.
129
Brynt Parmeter, Panelist: The Competitive Edge of Manufacturability. Panel at Mad Scientist Conference:
Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
130
Kira, op cit.
131
Kira, op cit.
132
Kira, op cit.
133
Sydney Freeberg, Breaking Defense. New Nuclear C2 Should be Distributed and Multi-Domain: STRATCOM
Deputy. April 05, 2017: http://breakingdefense.com/2017/04/new-nuclear-c2-should-be-distributed-multi-
domain-stratcom-deputy/
134
Evan Braden Montgomery, Center for Strategic and Budgetary Assessments, Reinforcing the Front Line: U.S.
Defense Strategy and the Rise of China (2017), p ii.
135
Army TRADOC G2, Mad Scientist 2016: The 2050 Cyber Army, (7 November 2016), p. 36.
136
Scharre, op cit.
137
Timothy A. Walton, Securing the Third Offset Strategy, Joint Forces Quarterly 83, 3 rd Quarter 2016.
138
Sadowski, op cit.
139
Sadowski, op cit.
140
Sadowski, op cit.
141
Scharre, op cit.
67
142
Scharre, op cit.
143
Scharre, op cit.
144
Scharre, op cit.
145
Sadowski, op cit.
146
The Joint Staff, Joint Concept for Robotics and Autonomous Systems (JCRAS), 19 October 2016.
147
The Joint Staff, Joint Concept for Robotics and Autonomous Systems (JCRAS), op cit.
148
Michelson, ibid.
149
Fountain, op cit.
150
Parmeter, op cit.
151
Ben FitzGerald, Alexandra Sander and Jacqueline Parziale, Center for New American Strategy Report, Future
Foundry: A New Strategic Approach to Military-Technical Advantage, 14 December 2016.
152
CNAS Report: Future Foundry: A New Strategic Approach to Military-Technical Advantage, ibid.
153
Parmeter, op cit.
154
Parmeter, op cit.
155
Parmeter, op cit.
156
Parmeter, op cit.
157
Kott, op cit.
158
Vincent Sabio, Program Manager at the Department of Defenses Strategic Capabilities Office, quoted by Todd
South in DOD Must Update How It Buys and Uses New Equipment, Technology for Future Battlefield, , Army
Times, May 2, 2017.
159
Kira, op cit.
160
Sadowski, op cit.
161
Sadowski, op cit.
162
Sadowski, op cit.
163
Sadowski, op cit.
164
Canton, op cit.
165
Canton, op cit.
166
Kira, op cit.
167
Andrew Ilachinski, op cit.
168
Pellerin, op cit.
169
Knight, op cit.
170
Andrew Ilachinski, op cit.
171
Andrew Ilachinski, op cit.
172
Andrew Ilachinski, op cit.
173
Andrew Ilachinski, op cit.
174
Fountain, op cit.
175
Kira, op cit.
176
Pippin, op cit.
177
Tom Simonite, AI Software Learns to Make AI Software, MIT Technology Review VOL 120 No 3, May /June
2017.
178
Sadowski, op cit.
179
Kott, op cit.
180
Aaron Chan, ASA(ALT). Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
181
Chan, op cit.
182
August Cole, nonresident Senior Fellow at the Brent Scowcroft Center on International Security at the Atlantic
Council and Director of The Art of the Future Project. Presentation to Mad Scientist Conference: Robotics, Artificial
Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050, 7 March 2017.
183
Cole, op cit.
184
Cole, op cit.
185
Cole, op cit.

68
186
Brian Michelson, the Atlantic Council Art of the Future Project, Warbot 1.0: The Death of Homer, February
2017; image by Alex Brady (Laser Tank) @ http://artoffuturewarfare.org/2017/02/warbot-1-0-the-death-of-
homer/
187
Thomas C Schelling in Foreword to Roberta Wohlstetters Pearl Harbor: Warning and Decision Stanford
University Press. 1962
188
Fountain, op cit.
189
Williamson Murray, Adaptation in War: With Fear of Change, Cambridge University Press, 2011.
190
Tom Greco, Closing Remarks,
191
Roper, Director, Strategic Capabilities Office. Quoted by Cheryl Pellerin in Defense Innovation Maintains
Military Overmatch Against Adversaries, DoD News, Defense Media Activity, 3 May 2017.

69
Appendix A: Workshop Design & Sources
Appendix A-1: Workshop Agenda

Mad Scientist 2017:

Robotics, Artificial Intelligence & Autonomy:
Visioning Multi Domain Battle in 2030-2050
7-8 March 2017
Georgia Tech Research Institute
Agenda Day 1: 7 March 2017

0800-0830 Registration

0830-0835 Administrative Remarks

Mr. Lee Grubbs, TRADOC

0835-0845 Technology Wargaming

Mr. Aaron Chan, ASA(ALT)

0845-0910 Welcome Remarks

LTG Kevin W. Mangum, DCG TRADOC

0910-0930 Opening Remarks

Dr Steve Cross, Executive Vice President for Research, Georgia Institute of
Technology

0930-0950 Dr. Augustus Fountain

Deputy Chief Scientist, Office of the Deputy Assistant Secretary of the Army
(Research & Technology)

0950-1010 Dr. Robert Sadowski

Robotics Senior Research Scientist Research, Technology and Integration
Director at U.S. Army TARDEC, U.S. Army Chief Roboticist

1010-1040 Break

1040-1140 Machine learning and artificial intelligence for sensor processing and perception
Dr. Zsolt Kira, Branch Chief of Advanced Machine Learning Analytics Group,
Robotics and Autonomous Systems Division at the GTRI

1140-1240 AI and Machine Learning/Translation

Dr. Jaime Carbonell, University Professor and Allan Newell Professor of Computer
Science at Carnegie Mellon

1240-1345 Lunch Break & Robot Petting Zoo

71
1345-1445 The Competitive Edge of Manufacturability
Brynt Parmeter, NetFlex

1445-1545 Death of the White Paper: How sci-fi stories, video games and film help us
understand and prepare for Future conflict
August Cole, nonresident Senior Fellow at the Brent Scowcroft Center on
International Security at the Atlantic Council and Director of The Art of the Future
Project

1545-1645 Unmanned Aerial Vehicle Swarms

Dr. Charles Pippin, Senior Research Scientist in the Aerospace, Transportation
and Advanced Systems (ATAS) Laboratory at GTRI

1645-1745 Arsenal of the Mind

Juliane Gallina, Director, Cognitive Solutions for National Security (North America)
IBM WATSON

1745-1800 Closing Remarks

MG Robert M. Dyess Jr., U.S Army, Deputy Director ARCIC

Agenda Day 2: 8 March 2017

0800-0830 Welcome Remarks

Mr. Thomas Greco, TRADOC G2

0830-0930 Convergence of Future Technology

Dr. James Canton, CEO and Chairman of the Institute for Global Futures

0930-1030 Robotics and Human/Robot Interaction

Dr. Magnus Egerstedt, Executive Director for the Institute for Robotics and
Intelligent Machines at the Georgia Institute of Technology

1030-1100 Break

1100-1200 Robotics and Sensors

Dr. Nahid Sidki, Executive Director of Robotics R&D, SRI International

1200-1300 Lunch Break

1300-1400 Unmanned and Autonomous systems

Paul Scharre, Senior Fellow and Director of the Future Warfare Initiative at the
Center for New American Security

72
1400-1500 Non-State actors and their uses of emerging technologies
Dr. Gary Ackerman, University of Maryland, National Consortium for the Study of
Terrorism and Responses to Terrorism

1500-1600 The Network is the Robot

Dr. Alexander Kott, Chief, Network Science Division, Computational and
Information Sciences Directorate, US Army Research Laboratory

1600-1700 Artificial Intelligence and Machine Learning: Potential Applications in Defense

Today and Tomorrow
Louis Mazziotta, Armament Research, Development, and Engineering Center

1700-1730 Closing Remarks

73
Appendix A-2: Conference Presenters

(IN ALPHABETICAL ORDER BY LAST NAME)

Dr. Gary Ackerman, University of Maryland, National Consortium for the Study of
Terrorism and Responses to Terrorism (Day 2)

Dr. James Canton, CEO and Chairman of the Institute for Global Futures (Day 2)

Dr. Jaime Carbonell, University Professor and Allan Newell Professor of Computer
Science at Carnegie Mellon (Day 1)

Aaron Chan, Office of the ASA(ALT) (Day 1)

August Cole, nonresident Senior Fellow at the Brent Scowcroft Center on International
Security at the Atlantic Council and Director of The Art of the Future Project (Day 1)

Dr Steve Cross, Executive Vice President for Research, Georgia Institute of Technology
(Day 1)

MG Robert M. Dyess Jr., U.S Army, Deputy Director ARCIC (Day 1)

Dr. Magnus Egerstedt, Executive Director for the Institute for Robotics and Intelligent
Machines at the Georgia Institute of Technology (Day 2)

Dr. Augustus Fountain, Deputy Chief Scientist, Office of the Deputy Assistant Secretary
of the Army (Research & Technology) (Day 1)

Juliane Gallina, Director, Cognitive Solutions for National Security (North America) IBM
WATSON (Day 1)

Mr. Thomas Greco, TRADOC G2 (Day 1&2)

Dr. Zsolt Kira, Branch Chief of Advanced Machine Learning Analytics Group, Robotics
and Autonomous Systems Division at the GTRI (Day 1)

Dr. Alexander Kott, Chief, Network Science Division, Computational and Information
Sciences Directorate, US Army Research Laboratory

LTG Kevin W. Mangum, DCG TRADOC (Day 1)

Louis Mazziotta, Armament Research, Development, and Engineering Center

Brynt Parmeter, NetFlex (Day 1)

Dr. Charles Pippin, Senior Research Scientist in the Aerospace, Transportation and
Advanced Systems (ATAS) Laboratory at GTRI (Day 1)

75
Dr. Robert Sadowski, Robotics Senior Research Scientist Research, Technology and
Integration Director at U.S. Army TARDEC, U.S. Army Chief Roboticist (Day 1)

Paul Scharre, Senior Fellow and Director of the Future Warfare Initiative at the Center for
New American Security (Day 2)

Dr. Nahid Sidki, Executive Director of Robotics R&D, SRI International (Day 2)

76
Appendix A-3: Conference Presentations

(IN ORDER OF PRESENTATION)

DAY ONE, 7 March 2017

Welcome Remarks
Mr. Lee Grubbs, TRADOC

Technology Wargaming
Mr. Aaron Chan, ASA(ALT)

Welcome Remarks
LTG Kevin W. Mangum, DCG TRADOC

Opening Remarks
Dr. Steve Cross, Executive Vice President for Research, Georgia Institute of Technology

Introductory Remarks
Dr. Augustus Fountain, Deputy Chief Scientist, Office of the Deputy Assistant Secretary
of the Army (Research & Technology)

Army Robotics Technology, Research, and Integration

Dr. Robert Sadowski, Robotics Senior Research Scientist Research, Technology and
Integration Director at U.S. Army TARDEC, U.S. Army Chief Roboticist

Machine Learning and Artificial Intelligence for Sensor Processing and Perception
Dr. Zsolt Kira, Branch Chief of Advanced Machine Learning Analytics Group, Robotics
and Autonomous Systems Division at the GTRI

AI and Machine Learning/Translation

Dr. Jaime Carbonell, University Professor and Allan Newell Professor of Computer
Science at Carnegie Mellon

The Competitive Edge of Manufacturability

Brynt Parmeter, NetFlex

Death of the White Paper

August Cole, nonresident Senior Fellow at the Brent Scowcroft Center on International
Security at the Atlantic Council and Director of The Art of the Future Project

Unmanned Aerial Vehicle Swarms

Dr. Charles Pippin, Senior Research Scientist in the Aerospace, Transportation and
Advanced Systems (ATAS) Laboratory at GTRI

77
Arsenal of the Mind
Juliane Gallina, Director, Cognitive Solutions for National Security (North
America) IBM WATSON

Closing Remarks
MG Robert M. Dyess Jr., U.S Army, Deputy Director ARCIC

DAY TWO, 8 March 2017

Welcome Remarks
Mr. Thomas Greco, TRADOC G2

Convergence of Future Technology

Dr. James Canton, CEO and Chairman of the Institute for Global Futures

Robotics and Human/Robot Interaction

Dr. Magnus Egerstedt, Executive Director for the Institute for Robotics and Intelligent
Machines at the Georgia Institute of Technology

Robotics and Sensors

Dr. Nahid Sidki, Executive Director of Robotics R&D, SRI International

Unmanned and Autonomous Systems

Paul Scharre, Senior Fellow and Director of the Future Warfare Initiative at the Center f
or New American Security

Non-State Actors and their Uses of Emerging Technologies

Dr. Gary Ackerman, University of Maryland, National Consortium for the Study of
Terrorism and Responses to Terrorism

The Network is the Robot

Dr. Alexander Kott, Chief, Network Science Division, Computational and Information
Sciences Directorate, US Army Research Laboratory

Artificial Intelligence and Machine Learning: Potential Applications in Defense Today and
Tomorrow
Louis Mazziotta, Armament Research, Development, and Engineering Center

Closing Remarks
Mr. Thomas Greco, TRADOC

78
Appendix B: SciTech Crowd-Sourcing Insights

SciTech Crowdsourcing Exercise Design

From 6-19 March, 2017, Office of the Deputy Assistant Secretary of the Army
(Research & Technology) in partnership with the U.S. Army Training and Doctrine
Commands (TRADOC) Mad Scientist Initiative partnered in a SciTech Futures online
crowdsourced exercise.

Crowdsourcing is the practice of obtaining information or input into a task or project by enlisting
the services of a large number of people, either paid or unpaid, typically via the internet.

The term crowdsourcing was popularized in a 2004 book by James Surowieki called The Wisdom of
Crowds. In it, he described an experiment in 1906 by Francis Galton in which over 800 people asked
to judge the weight of an Ox. The average guess was very close to the actual weight. In the internet
age, the ability of a great many to collaborate and refine a single idea has allowed crowdsourcing to
become a powerful tool for online prediction markets and if well designed, high quality forecasts that
are better on average than individual expert opinion.

This exercise consisted of three basic parts. The Imaginarium was a virtual space in
which online participants could post ideas and brainstorm issues related to robotics, AI
and autonomous systems. The Workshop was a place in which promising ideas as
decided by exercise participants could be sharpened and refined collaboratively. The
third part of crowdsourcing exercise was the Marketplace. In it, participants invested
virtual currency in promising ideas.
Each registered exercise participant was given a fixed, non-transferrable amount of
virtual currency each day and could place that money in ideas they believed held the
most promise. Participants could evaluate ideas based on individual criteria found on
each idea page, which included:
A description of the idea.
Description of special characteristics that make each idea new or different.
Implications of the idea.
Top improvements to the idea by other collaborators.
Time frame when the idea will make a difference.
Who will benefit most from the idea, for example, the Army, society, adversaries,
etc.
The SciTech Futures Crowdsourcing Exercise web site is at:
https://scitechfutures.com/ex6/.

79
Leading Crowdsourced-Developed Ideas
The nearly two-week exercise resulted in 139 distinct, investible ideas within the
Marketplace. Crowdsource exercise participants invested in 134 of the 139 ideas that
transitioned to the marketplace. The average investment in any particular idea was
$9500. The top thirty ideas by amount of investment dollars over the course of the
exercise were as follows:
Amount Year
Rank Idea Name
Invested Achieved
1 Exponential Human Intelligence $67,000 2045
2 Man vs. Machine Enterprise Level Approach $55,000 2020
3 Smart Dust $41,000 2030
4 Machine Augmentation to Staff Functions $33,000 2025
5 Crowd-Source Game Environment $30,500 2020
6 Recon-By Wire $28,000 2025
7 EMP Protection Inc. $25,500 2035
8 Robotics/AI/Cyber Geneva Convention $23,000 2030
9 Machine Learning Pathologies $21,500 2035
10 AI Planetary Colonization $21,000 2050
11 Motorpool Bots $20,500 2030
12 AI-Enhanced Network Gatekeepers $19,000 2020
13 DigiPatton $18,000 2040
14 Ten-Cent Defeat $17,500 2020
15 Robotic CASEVAC $17,000 2025
16 Mesh Networks as Alternate Internet $16,500 2030
17 Combat Engineer Bots $16,000 2030
18 Combating Space Junk $15,500 2035
19 Additive Manufacturing Sustainment Brigades $15,000 2025
20 (tie) Follow Me! $14,500 2030
20 (tie) Nano-AI Vaccinations $14,500 2030
22 (tie) Go Medieval $14,000 2025
22 (tie) Potential Replacements for Honey Bees $14,000 2040
22 (tie) Soft Robotics for Triage $14,000 2030
25 (tie) Robotic Wingman $13,500 2030
25 (tie) FC-48 Fabship Aircraft $13,500 2025
27 Telecommuting to War $13,000 2035
28 Robo Lingo $12,500 2030
29 (tie) Multi-Layer Multi-Spectral Lens Protection $12,000 2020
29 (tie) Anti-Pattern Recognition Camo $12,000 2030
29 (tie) Counter-AI Operations Field Manual $12,000 2030
29 (tie) Mr. Trashwheel for Space $12,000 2040
30 Recon Round $11,500 2030

Marketplace Ideas by OE Solution Approach

Each of the ideas were judged by participants as to the most likely year they would be
achieved or have an effect. These judgements were averaged, and the graphic below

80
provides and overall picture of the spread of when individual ideas would most likely
be available. Most ideas clustered in the 2030 time frame.

After completion of the exercise, the research team collected each marketplace ideas
and arranged them by year according to the Operating Environment solution elements
in Section III of this technical report. These are:
Manned-Unmanned Teaming
Asymmetric Awareness and Decision
Swarming
Intelligent Networks for the Internet of Battle Things
Autonomous Sustainment
Post exercise analysis arranged each of the ideas by relevance to the OE thematic
areas by the time they were expected to be available. Top 30 ideas within each area
were marked by red double asterisks (**).
Several highly ranked ideas within the top 30 (for example, Exponential Human
Intelligence, Robotics/AI/Cyber Geneva Convention or Exponential Human Intelligence)
could not be easily categorized or were irrelevant to the OE solution elements, and are
not found in the solution element/crowdsourced idea tables.
MUM-T Crowdsourcing Observations. MUM-T-related ideas accounted for some 13%
(19 of 139) of the total ideas developed during the crowdsourcing exercise. Of these,
four ideas were represented in the top 30 ideas relevant to the Army in the future
(marked with ** in the table). The top idea, Man vs. Machine: Human, AI, and Robotic
Employment Optimization was the second highest-rated idea in the entire exercise.
This idea envisions a DOD enterprise-wide approaches to evaluating the relative merits

81
of human labor vice machine labor within the DoD. This idea promises a structured
method to leverage the disruptive effect of robotics, AI, and autonomy and apply MUM-
T and machine labor in an optimal way. It was imagined to be available to the force by
exercise participants by 2020.

The full set of MUM-T crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Manned Unmanned Teaming: Idea Name
Achieved
2020 Man vs. Machine: Human, AI, and Robotic Employment Optimization**
2020 Sort Robotic Wearables: Autonomous Tourniquet
2020 Autonomous Mine Removal
2025 Mobile Protected Firepower
2025 Cyber Manifesto
2025 AWACS 3.0 Distributed Robotic Battle Management
2025 Appliqu Autonomy Kits
2025 AI-Assist for Combat Medic
2030 Walking Combat Vehicles
2030 Robotic Wingman**
2030 Combat Engineer Bots**
2030 Combat Robot Ethos
2035 Telecommuting to War**
2035 Second Skin
2035 Backup Brain
2040 Human Accessible Robot/AI Off Switch
2040 Remote Operated Assault Robots
2040 Enhanced Others
2045 Insect Man

Asymmetric Awareness and Decision Crowdsourcing Observations. AA&D

related ideas represented the single most popular category for ideas during the
exercise, accounting for nearly 24% (33 of 139) of the total set. Eight of these ideas
were represented in the top 30 ideas relevant to the Army in the future (marked with **
in the table). The top idea, Machine Augmentation to Staff Functions was the fourth
highest-rated idea in the entire exercise. This idea described the application of deep
learning and AI to optimize thousands of course of action, from logistics, to personnel,
intelligence, and operations fed from thousands of sensor and ISR feeds. This idea
would allow staffs to operate within adversary decision cycles and allow the Army to
seize and retain the initiative during operations. This idea was imagined to be available
to the force by exercise participants by 2025.

82
The full set of AA&D crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Asymmetric Awareness and Decision: Idea Name
Achieved
2020 Multi-Layer Multi-Spectral Lens Protection**
2020 Real News Aggregator
2020 A.I Assisted Searchable Portable Military Library Laptop
2020 Pocket Augmented Reality Real-Time Training
2025 Adversaries Simulating Us
2025 Autonomous Sensor Defeat
2025 Heads-Up Glasses, Dash, and Desk Displays
2025 Pocket Interactive Doctrine, Training, and Policies
2025 Anti-Autonomy Sensor Disruptors
2025 Military/Law Enforcement Rehearsals
2025 Kinetic Courier / Kinetic Jammer
2025 Multi-Mode Laser Designator
2025 Machine Augmentation to Staff Functions**
2025 Robotic Subterranean Operations
2025 AI Robotic Information Warriors
2025 Adaptive Hyperspectral Algorithm for Camouflage Detection
2025 Recon-by-Wire**
2030 Chatbot: AI Resurrected Clones of Great Thinkers
2030 EW Applied to Human Perception
2030 Cybernetic Super-Sniffers
2030 Misinformation Disintegrator
2030 Anti-Pattern Recognition Camo**
2030 Mesh Networks as Alternate Internet**
2030 Recon Round**
2030 Genetic Algorithms for Optimizing Team Composition
2030 Rent-an-Avatar Booth
2030 Counter-AI Operations Field Manual**
2035 Second Skin
2035 21st Century Non-Kinetic, Multidomain Training for All Troops
2035 TOC in a Box
2035 Ever-Present Commander Rules of Engagement Authority
2040 DigiPatton**
2045 Ultra-Fast Battlefield

83
Swarming Crowdsourcing Observations. Swarming-related ideas accounted for 8%
(12 of 139) of the total set. Two of these ideas were represented in the top 30 ideas
relevant to the Army in the future (marked with ** in the table). The top idea, Ten Cent
Defeat was the 14th highest-rated idea in the entire exercise. This idea described the
ability of all robots and autonomous systems to not fail spectacularly when confronted
with primitive, low-cost defeat mechanisms, adapt, and recover functionality. The idea
would apply a range of technologies and approaches to ensure that some percentage of
a robotic fleet would remain operational even when confronted with novel
countermeasures. This idea imagined to be available to the force by exercise
participants by 2020.

The full set of Swarming crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Swarming: Idea Name
Achieved
2020 Ten-Cent Defeat**
2025 Virtual Minefield
2025 Drone Swarms
2025 Mothership/UCAV Delivery Carrier
2030 Nano-AI Vaccines**
2035 AI Prototype Platform Design
2035 Autonomous Infrastructure Repair and Maintenance
2035 Swarming Attack Nano-Bots
2035 Permanent Protective Drone Swarms
2035 Nanobot/Microbot Tracing Sensors
2040 Sleeper Drones
2045 Attack of the Clones

Intelligent Networks for the Internet Battle of Things Crowdsourcing

Observations. Intelligent network-related ideas accounted for 13% (18 of 139) of the
total set. Three of these ideas were represented in the top 30 ideas relevant to the
Army in the future (marked with ** in the table). The top idea, Smart Dust was the third
highest-rated idea in the exercise. This idea described radio-frequency identification
(RFID) transmitters the size of a human hair with unique number strings for tracking
purposes which are deployed in varying amounts for discrete or mass surveillance. The
idea would provide a new range of ISR capabilities to the force to track and monitor
targets remotely and with high quality data. This idea imagined to be available to the
force by exercise participants by 2030.

The full set of Intelligent Networks crowdsourced ideas are as follows:

84
 Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Achieved
Intelligent Networks for the Internet of Battle Things: Idea Name
2020 AI-Enhanced Network Gate-keepers**
2020 AI Research Assistants
2020 Multi-Function Weapons
2020 Report Writer: Customizable AI Research Tool
2020 Robotic Programmers, Inc.
2025 Plug and Play Military Robotics Vehicles
2025 Corrupted R&D Simulation Software
2025 Kit to Control Captured Enemy Equipment
2025 Disrupter Bots for Crowd-Sourced Online Studies
2025 Algorithms to Approximate Human Judgments
2030 Teams of Small Robots to Move Casualties to Safety
2030 Anti-Machine Pathogens
2030 Smart Dust**
2030 AI Overrun Protection
2030 Neuronet
2030 Internet of (Hostile) Things
2030 Networked Autonomous Infrastructure Sabotage Battalion
2035 Machine Learning Pathologies**

Autonomous Sustainment Crowdsourcing Observations. Autonomous

Sustainment-related ideas accounted for 8% (12 of 139) of the total set. Four of these
ideas were represented in the top 30 ideas relevant to the Army in the future (marked
with ** in the table). The top idea, Motorpool Bots was the 11th highest-rated idea in the
exercise. This idea described need to develop a capability to repair and maintain robots
in the future. Robots may significantly enhance PMCS as well as perform repairs and
system upgrades. Once they master the controlled environment, these systems could
then be outfitted with cross-country terrain mobility systems so they can follow units into
the field, repairing and recovering damaged system even under direct or indirect fire.
The idea would allow robots to undertake dirty, dull, and dangerous repair tasks for the
Army. This idea imagined to be available to the force by exercise participants by 2030.

The full set of Autonomous Sustainment crowdsourced ideas are as follows:

Mad Scientist SciTech Futures Crowd-Sourcing Exercise

Year
Achieved
Autonomous Sustainment: Idea Name
2025 3D Printing for Maintenance Parts

85
2025 Additive Manufacturing Sustainment Brigades**
2025 Hoarder Drone
2025 Fabship Aircraft**
2025 Six Sigma Army Total Design and Maintenance
2025 Robotic CASEVAC**
2030 AI Based New Product Development
2030 Motorpool Bots**
2035 Walking Emergency or Construction Vehicles
2035 Integrated Electrical Logistics
2045 Autonomous Space Miners
2045 BN/BDE Experimentation and Upgrade Officer

86
Appendix C: Collection and Assessment Methodology
This appendix describes the collection, organization, and assessment of data,
information, and knowledge for the Mad Scientist 2017 Conference: Robotics,
Artificial Intelligence, and Autonomy in Multi-Domain Battle 2030-2050, including
associated papers, speaker notes, conference discussions, and observations derived
from participants in a parallel crowd sourced wargame (known as the SciTech Futures
Exercise). Our overall approach to data collection and analysis is captured in Figure 1
below.

Collection and Assessment Description

Topic
Background What is the situation being studied?
Purpose Why is this study being conducted?
Key Tasks What tasks must be accomplished, and who will do them?
End State and Deliverables What will this effort produce? What is the deadline for the
project?
Scope What are the limits of this collection effort? Who will be involved?
Concept What is the scale of effort and what areas must be examined?
Who will conduct the study? What is the time frame for the study?
Research Questions What are the issues to be examined? What questions must be
asked to examine those issues? Optionally, hypothesize what
you are trying to confirm or deny.
Key Personnel and Organizations Who can answer these questions? Develop a list of key
personnel to be interviewed.
Methodology How will the study be organized? How will various teams
interface?
Reference Material What will be the primary documents of reference? How will they
be applied in the study?
Data Collection Procedures What quantitative and qualitative data must be collected, and
how and when?
Data Management Procedures How will collected data be managed? Who will have access to
the data and at what stages of collection and analysis? Who has
release authority? What are the classification procedures?
Figure 1: Overall Approach to Data Collection and Assessment

Background
On 7-8 March, United States Army TRADOC G2 conducted Mad Scientist 2017
Conference: Robotics, Artificial Intelligence, and Autonomy in Multi-Domain Battle 2030-
2050 in partnership with the Georgia Tech Research Institute. This event explored the
Armys robotics, AI, and autonomy requirements for multi-domain battle in the 2035-
2050 period. This conference is part of a larger United States Army TRADOC Mad
Scientist Series in support of the overall Army Campaign of Learning.

Purpose
87
The Mad Scientist 2017 Conference: Robotics, Artificial Intelligence, and Autonomy in
Multi-Domain Battle 2030-2050 was designed to support the broader Army Mad
Scientist initiative goals to continuously adapt, innovate, and allow for broader
engagement in problem solving within the far future of armed conflict. This conference
explored two broad themes related to the application of robotics, AI, and autonomous
systems in future warfare. These were:
How will AI and robotics change the relationship between humans and warfare?
How could AI, robotics, and autonomous systems enable the United States
military and its adversaries in multi-domain warfare?
This collection and assessment methodology describes how the analysis team collected
and assessed the event data to provide observations and insights captured in the
Quicklook and Technical Reports.
Key Tasks
Key tasks for this collection and assessment effort were derived from the Robotics, AI,
and Autonomy Information Paper and the GTRI/Mad Scientist Primer briefing and
included:
Prepare for the Robotics, AI, and Autonomy Conference by developing a
collection and assessment protocol (due 6 March 2017).
Develop an analytic method to produce a Technical Report (due 6 March 2017).
Observe briefings and panel discussions during the event, collect detailed notes
on conference proceedings, and gather and organize the results of each phase
of the conference.
Collect and assess ongoing activities and responses within the SciTech Futures
Exercise.
Assess the results of the Robotics, AI, and Autonomy Conference.
Generate a Quicklook Report providing initial insights for the TRADOC G2 AAR
process (due 7 April 2017)
Produce a Technical Report with the results of the Robotics, AI, and Autonomy
Conference that further refines our understanding of these issues to effectively
support regional, global, joint and Army Operations in Multi-Domain Battle, 2030-
2050, as well as those capabilities a potential adversary may employ.
Deliver a Technical Report within no later than 19 May 2017)
Support HQ TRADOC analytical team, by collecting notes and developing
observations and insights during the event and from live stream questions and
comments and providing consolidated insights to forward TRADOC G-2
personnel at event, to aid in updates and briefings to senior U.S. Army
personnel.
End State and Deliverables

88
The observations and insights generated in accordance with this collection and
assessment methodology enabled the delivery of key insights for senior Army leaders to
support the assessment of technical challenges and potential solutions related to the
use of robotics, AI, and autonomy in the context of Multi Domain Battle. This analytic
report is intended to assist in TRADOC G-2s understanding of these issues out to 2050.
This report provides provide additional focus on how AI and robotics may change the
relationship between humans and warfare. Insights are also designed to support the
Army campaign of learning, contributing to an improved understanding of conflict and
the character of war in the future operating environment.
All observations and insights were collected, refined, and presented in a Quicklook
Report (delivered 7 April 2017) and a Technical Report (delivered 19 May 2017) that
provide initial and refined observations and insights consolidating relevant data from the
conference presentations, SciTech Futures Exercise, and other associated data.
The findings of the Technical Report address five potential solution areas which
emerged as dominant themes during the conference. These included Manned-
Unmanned Teaming, Asymmetric Awareness and Decision, Swarming, Intelligent
Networks for the Internet of Battle Things, and Autonomous Sustainment. Section III of
the Technical Report provides a view of potential robotics, AI, and autonomy solutions
within each of these themes.
Additionally, the Technical Report arranges ideas developed during the SciTech Futures
Exercise along these five themes. Each idea is arranged according to the year that
participants believed that the solution might be available to the Army. A full accounting
of the ideas uncovered during the exercise are found at appendix B of this report.
Observations and insights were further explored in terms of how they may impact nine
Competitions of Multi-Domain Warfare derived from a draft study The Operational
Environment, 2035-2050: The Emerging Character of Warfare. These competitions
include:
Finders vs Hiders
Strikers vs Shielders
Range & Lethality vs Close Engagement & Survivability
Disconnection / Disaggregation / Decentralization vs Connection / Aggregation /
Centralization
Offense vs Defense
Planning & Judgement vs Reaction & Autonomy
Escalation vs De-Escalation
Domain vs Domain
Dimension vs Dimension

Scope

89
This collection and assessment methodology describes how the team captured and
refined the data, information, and knowledge developed for and during the March 2017
Robotics, Artificial Intelligence, and Autonomy in Multi-Domain Battle 2030-2050
Conference, as well as robotics, AI, and autonomy-related policies and strategies as
well as relevant studies that explore the nature of these advanced technologies as well
as their impact on the future of warfare.
Concept
The concept to collect and assess information generated over the course of the
Robotics, Artificial Intelligence, and Autonomy in Multi-Domain Battle 2030-2050
Conference included the following elements:
Conduct a survey of studies, reports, or concepts related to robotics, AI, and
autonomy.
Develop analytic structure for Technical Report that relates to Army S&T lines of
effort.
Collect notes from the assessment team captured over the course of the
Robotics, Artificial Intelligence, and Autonomy in Multi-Domain Battle 2030-2050
using a structured set of information elements related to each of the research
questions (see Research Questions and Methodology below)
Observe SciTech Futures Exercise research questions, discussions, and outputs
and integrate observations into Quicklook and Technical Reports.
Assess the results of the Robotics, Artificial Intelligence, and Autonomy in Multi-
Domain Battle 2030-2050 Conference, and write an initial insights (Quicklook)
report by 7 April 2017.
Write a Technical Report with the results of the conference, with specific
recommendations to the impact of these capabilities (both U.S. and adversary)
on Multi-Domain Battle.
Complete and delivery Technical Report by 19 May, 2017.

Research Questions
This research effort was guided by seven research questions that focused note-taking,
continuous analysis, and observations and insights development. These questions
included:
1. What roles will AI/machine learning have in the planning, preparing, and
execution of combat operations in a multi-domain conflict?
2. How will robotics and autonomy change the roles of Soldiers and Leaders in
future combat?
3. How will AI/machine learning help Leaders visualize combat operations ongoing
across all domains?

90
 4. What are the ethical considerations and vulnerabilities of using or not using
autonomous systems in lethal operations?
5. What are the possibilities for human machine interface that will allow Leaders to
offload mental and physical responsibilities?
6. How could adversaries gain tactical and operational advantages over U.S.
military forces using AI and autonomous devices?
7. What other trends will be greatly impacted accelerated or amplified with the
explosion of AI and autonomy in society and war?

Key Personnel
The analytic effort was undertaken by Mr. David Fastabend, and Mr. Jeff Becker. This
effort also relies on close collaboration with others key partners in the broader effort
throughout the analytic effort, including:
Overall study integration and senior leader support: Mr. Tom Schmidt, TRADOC
G2
TRADOC G2 POC: Mr. Lee Grubbs, TRADOC G2
SciTech Futures Exercise Observations: Mr. Luke Shabro, TRADOC G2

Methodology
The methodology to assess data and information collected over the course of the
Robotics, Artificial Intelligence, and Autonomy in Multi-Domain Battle 2030-2050
Conference occured over the following four phases.
In phase 1 (pre-conference preparation), the team conducted a comprehensive review
of applicable literature, including prior Mad Scientist study reports, as well as reference
material cited at Appendix C. The team began formulating the Quicklook and Final
Report structures by examining the overarching themes and associated research
questions, and developing a report structures that communicate key ideas such that
they are easily relatable to Army S&T priorities (see Appendix A).
In phase 2 (conference execution), two members of the research team (Fastabend;
Becker) were located on-site and attend all Conference proceedings. The conference
was designed around briefings by featured speakers explore issues or topics related to
robotics, AI, and autonomous systems. Each was intended to spark discussion among
conference participants about how these technologies and capabilities may evolve out
to 2050 and the implications of these changes for the future of warfare but particularly
for the Army and Multi-Domain Battle.
The note-taking and observation development team conducted continuous assessment
and synthesis of the proceedings. In order to capture conference presentations and
discussion sufficient to address the research questions, the team took detailed notes
and conducted continuous assessment based on the seven research questions outlined
above. Note takers met meet each day immediately before and after conference

91
sessions to share observations, and provided notes each day to Mr. Fastabend and Mr.
Becker.
The team listened to each speaker presentation, and collected notes based on this
method. As necessary, the team engaged with conference participants both during and
after the conference to further refine and develop ideas. The team collected briefings
for reference during phase 3 of the methodology. The team integrated any written
materials from these panels and briefings as the foundation the Quicklook and
Technical Report development and writing efforts as well.
In phase 3 (Quicklook Report development) the team developed an initial synthesis of
key findings related to the research questions. The Quicklook Report was organized
according to several broad thematic areas, and focused on surfacing and refining
important issues described during the Robotics, Artificial Intelligence, and Autonomy in
Multi-Domain Battle 2030-2050 Conference presentations and proceedings, as well as
the SciTech Futures Exercise results. The exercise took place concurrently with and
after the conference (from 6-19 March) and was monitored after the exercise was
completed.
The team delivered a 2050 Robotics, Artificial Intelligence, and Autonomy in Multi-
Domain Battle 2030-2050 Quicklook Briefing in Microsoft Word format. It describe
emerging themes in order to support AAR development for the wider TRADOC G2
effort.
In phase 4 (Technical Report Construction) the team constructed a Technical Report
keyed to inform overall TRADOC OE development and the Army Campaign of Learning.
This report summarized key observations and insights from the event, the SciTech
Futures Exercise, as well as implications of findings derived from the prior three phases
of the study.
Reference Material

Primary reference material associated with the study is cited at Appendix D. The
reference material list evolved throughout the study process and is provided as part of
this final report.

Data Collection Procedures

The team conducted real-time collection management to ensure the collection of
accurate and complete impressions of the event, and that notes are shared between
both team members. Our analysts have documented notes from each panel and
presentation of a summary sheet in Microsoft Word. Each set of notes was be
collected, integrated, and stored on a Microsoft OneDrive shared file system in a
Microsoft Word file. These summaries were also be shared and saved on two
independent computers for continuity of operations.

92
The notes and analysis team held daily collaboration sessions to share key insights
from the days work and to begin to identify key and recurring themes. This disciplined
and methodical cataloguing of summaries and other documents, coupled with the verbal
discourse during the event enabled our team to analyze conference proceedings and
develop observations and insights in a timely manner for the Quicklook Report, and
comprehensively for the Technical Report.
Data Management Procedures
Data collected during the event was managed individually be the team members. The
information was shared via Google Gmail and Microsoft OneDrive file structures. Only
note and analysis team members have access to the data. Data release was managed
by Mr. Fastabend, who will provide TRADOC G2 raw collected data and analytic
materials at their request. This material is unclassified, but until publically released, is
sensitive in nature. As such, it will not be shared except between the team members
and between the team and TRADOC G2 authorities.

93
THIS PAGE INTENTIONALLY LEFT BLANK

94
Appendix D: References

Ackerman, Dr. Gary. University of Maryland, National Consortium for the Study of
Terrorism and Responses to Terrorism. Presentation to Mad Scientist Conference:
Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare
in 2030-2050 8 March 2017.
APMAS 2011 Conference Website. Revolutionary New Materials! @
http://www.apmas2011.org/revolutionary-new-materials.html
Brianwieck. TRADOC G2 Mad Scientist SciTech Crowd-Sourcing Exercise @
https://scitechfutures.com/ex6/workshop/ai-robotic-information-warriors/
Brissett, Wilson. Not Enough People to Solve the Cyber Threat Air Force
Magazine Daily Report, 4 May 2017
Brooks, Rosa. Can There Be War Without Soldiers? Foreign Policy Magazine, 15
March 2016.
The Burgess Shale Website The Cambrian Explosion @ http://burgess-
shale.rom.on.ca/en/science/origin/04-cambrian-explosion.php#unique
Byrnes, Nanette. Goldman Sachs Embraces Automation, Leaving Many Behind
MIT Technology Review VOL 120 No 3, May /June 2017.
Caddell, Cate. China's Baidu buys U.S. computer vision startup amid AI push Reuters
Technology News, 13 Apr 2017 @ http://www.reuters.com/article/us-baidu-m-a-
idUSKBN17F0JF
Canton, Dr James. CEO and Chairman of the Institute for Global Futures.
Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050 8 March 2017.
Carbonell, Dr. Jaime. Active and Proactive Machine Learning: From Fundamentals
to Applications in Computational Biology, Machine Translation and Wind Energy
Brigham Young University Computer Science Website @
https://cs.byu.edu/colloquium/active-and-proactive-machine-learning
Carbonell, Dr. Jaime. University Professor and Allan Newell Professor of Computer
Science at Carnegie Mellon. Presentation to Mad Scientist Conference: Robotics,
Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-
2050 7 March 2017.
Chan, Aaron. Office of the ASA(ALT). Presentation to Mad Scientist Conference:
Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare
in 2030-2050 7 March 2017.
Cole, August. Nonresident Senior Fellow at the Brent Scowcroft Center on International
Security at the Atlantic Council and Director of The Art of the Future Project.

95
Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050 7 March 2017.
Defense Science Board Summer Study on Autonomy June 2016.
Egerstedt, Dr. Magnus. Executive Director for the Institute for Robotics and Intelligent
Machines at the Georgia Institute of Technology. Presentation to Mad Scientist
Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-
Domain Warfare in 2030-2050 8 March 2017.
FitzGerald, Ben; Sander, Alexandra and Parziale, Jacqueline. Center for New
American Strategy Report. Future Foundry: A New Strategic Approach to Military-
Technical Advantage 14 December 2016.
Fountain, Dr. Augustus. Deputy Chief Scientist (ST) Office of the Deputy Assistant
Secretary of the Army (Research & Technology). Presentation to Mad Scientist
Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-
Domain Warfare in 2030-2050. 7 March 2017.
Freeberg, Sydney. Breaking Defense. New Nuclear C2 Should be Distributed and
Multi-Domain: STRATCOM Deputy April 05, 2017: @
http://breakingdefense.com/2017/04/new-nuclear-c2-should-be-distributed-multi-
domain-stratcom-deputy/
Gallina, Juliane. Director, Cognitive Solutions for National Security (North America) IBM
WATSON. Presentation to Mad Scientist Conference: Robotics, Artificial
Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050 7
March 2017.
Greco, Tom. Opening Remarks to the Mad Scientist Conference: Robotics,
Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-
2050 7 March 2017.
Greco, Tom. Closing Remarks to the Mad Scientist Conference: Robotics,
Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-
2050 8 March 2017.
Hambling, David. The Next Era of Drones will be Defined by Swarms BBC, 27
April 2017.
Harper, Jon. Army, Industry See Bright Future For Robotic Vehicles National
Defense Magazine, 19 April 2017.
Horowitz, Michael. Foreign Policy Magazine. The Looming Robotics Gap 5 May
2014, @ http://foreignpolicy.com/2014/05/05/the-looming-robotics-gap/
Ilachinski, Andrew. Center for Naval Analysis. AI, Robots, and Swarms: Issues,
Questions, and Recommended Studies, January 2017.
Institute of the Future Web Site. Blended Reality: Superstructing Reality,
Superstructing Selves http://www.iftf.org/our-work/people-technology/technology-
horizons/blended-reality/

96
The International Weekly Journal of Science, nature website. What Caused the
Cambrian Explosion? @ http://www.nature.com/news/what-sparked-the-cambrian-
explosion-1.19379CNAReport
The Joint Staff. Joint Concept for Robotics and Autonomous Systems (JCRAS)
19 October 2016.
Kelly, Kevin. The Three Breakthroughs That Have Finally Unleashed AI on the
World Wired Magazine, 27 October 2014.
Kira, Dr. Zsolt. Branch Chief of Advanced Machine Learning Analytics Group within the
Robotics and Autonomous Systems Division at the Georgia Tech Research Institute.
Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050 7 March 2017.
Knight, Will. The Dark Secret at the Heart of AI MIT Technology Review VOL 120
No 3, May /June 2017.
Kott, Dr Alexander. Chief, Network Science Division, Computational and Information
Sciences Directorate, US Army Research Laboratory. Presentation to Mad Scientist
Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning Multi-
Domain Warfare in 2030-2050 8 March 2017.
Libicki, Martin. The Mesh and the Net: Speculation on Armed Conflicts in an Age
or Free Silicon. McNair Paper 28 (Washington, D.C.) 1996.
Montgomery, Evan Braden. Center for Strategic and Budgetary Assessments.
Reinforcing the Front Line: U.S. Defense Strategy and the Rise of China 2017.
Parmeter, Brynt. Panelist: The Competitive Edge of Manufacturability. Panel at Mad
Scientist Conference: Robotics, Artificial Intelligence and Autonomy: Envisioning
Multi-Domain Warfare in 2030-2050 7 March 2017.
Pippin, Dr. Charles. Senior Research Scientist in the Aerospace, Transportation and
Advanced Systems (ATAS) Laboratory at Georgia Tech Research Institute.
Presentation to Mad Scientist Conference: Robotics, Artificial Intelligence and
Autonomy: Envisioning Multi-Domain Warfare in 2030-2050 7 March 2017.
Pratt, Gil. Is a Cambrian Explosion Coming for Robotics? IIEE Spectrum (31
August 2015).
Mazziotta, Louis. Armament Research, Development, and Engineering Center
(ARDEC). Presentation to Mad Scientist Conference: Robotics, Artificial
Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-2050 8
March 2017.
Michelson, Brian M. The Bridge. Blitzkrieg Redux: The Coming Warbot
Revolution (6 Mar 2017)
Michelson, Brian. The Atlantic Council Art of the Future Project, Warbot 1.0: The
Death of Homer February 2017; image by Alex Brady (Laser Tank) @
http://artoffuturewarfare.org/2017/02/warbot-1-0-the-death-of-homer/

97
Muro, Mark; McAffee, Andrew; and Rotman, David in Letters to the Editor, MIT
Technology Review VOL 120 No 3, May /June 2017.
Murray, Williamson. Adaptation in War: With Fear of Change Cambridge University
Press. 2011.
National Public Radio. Planet Money Podcast: BOTUS Episode 763, April 7, 2017.
Pellerin, Cheryl. DoD News, Defense Media Activity. Defense Innovation Maintains
Military Overmatch Against Adversaries 03 May 2017 @
https://www.defense.gov/News/Article/Article/1172099/defense-innovation-maintains-
military-overmatch-against-adversaries/
Potember, Richard. Perspectives on Research in Artificial Intelligence and
Artificial General Intelligence Relevant to DoD JASON, The MITRE Corporation.
January 2017.
Roper, Dr Will. Director, Strategic Capabilities Office. Quoted by Cheryl Pellerin in
Defense Innovation Maintains Military Overmatch Against Adversaries DoD
News, Defense Media Activity, 3 May 2017.
Rotman, David Desktop Metal Thinks Its Machines Will Give Designers and
Manufacturers a Practicable and Affordable Way to Print Metal Parts MIT
Technology Review VOL 120 No 3, May /June 2017.
Sabio, Vincent. Program Manager at the Department of Defenses Strategic
Capabilities Office, quoted by Todd South in DOD Must Update How It Buys and
Uses New Equipment, Technology for Future Battlefield Army Times, May 2, 2017.
Sadowski, Dr Robert. Robotics Senior Research Scientist Research, Technology and
Integration Director at U.S. Army TARDEC, U.S. Army Chief Roboticist Presentation
to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy:
Envisioning Multi-Domain Warfare in 2030-2050 7 March 2017.
Scharre, Paul. Senior Fellow and Director of the Future Warfare Initiative at the Center
for New American Security. Centaur Warfighting: The False Choice of Humans vs
Automation Temple University website @
https://sites.temple.edu/ticlj/files/2017/02/30.1.Scharre-TICLJ.pdf
Scharre, Paul. Senior Fellow and Director of the Future Warfare Initiative at the Center
for New American Security. Presentation to Mad Scientist Conference: Robotics,
Artificial Intelligence and Autonomy: Envisioning Multi-Domain Warfare in 2030-
2050 8 March 2017.
Scharre, Paul. Senior Fellow and Director of the Future Warfare Initiative at the Center
for New American Security. Robotics on the Battlefield Part II: The Coming
Swarm CNAS website @ https://www.cnas.org/publications/reports/robotics-on-the-
battlefield-part-ii-the-coming-swarm
Schelling, Thomas C in Foreword to Roberta Wohlstetters Pearl Harbor: Warning
and Decision Stanford University Press. 1962

98
Sidki, Dr. Nahid. Executive Director of Robotics R&D, SRI International. Presentation
to Mad Scientist Conference: Robotics, Artificial Intelligence and Autonomy:
Envisioning Multi-Domain Warfare in 2030-2050 8 March 2017.
Simonite, Tom. AI Software Learns to Make AI Software MIT Technology Review
VOL 120 No 3, May /June 2017.
Stevens, Laura and Higgins, Tim. Amazon Studies Driverless Ideas The Wall
Street Journal (Technology) p B4, 25 April 2017.
Swanson, Bret and Mandel, Michael. Robots Will Save the Economy The Wall
Street Journal, 15 May.
Tucker, Patrick. The Military is Using Human Brain Waves to Teach Robots How
to Shoot Defense One, 5 May 2017.
U.S. Army. The Army Cyberspace Strategy for Unified Land Operations January
2016.
U.S. Army Training and Doctrine Command G2. Mad Scientist Conference:
Strategic Security Environment in 2025 and Beyond (October 2016).
U.S. Army Training and Doctrine Command G2. Mad Scientist Conference: The
2050 Cyber Army November 2016).
U.S. Army (Draft) Training and Doctrine Command G2 Paper: The Operational
Environment, 2035-2050: The Emerging Character of Warfare, March 2017.
U.S. Army Training and Doctrine Command, Maneuver Aviation, and Soldier Division
Army Capabilities Integration Center. Robotics and Autonomous Systems (RAS)
Strategy January 2017.
Vaganova, Elena V. Department of Innovative Technologies, National Research Tomsk
State University. Indicators of Innovation Potential of the Country as Means of the
Government Policy Modeling in the Dominant and Emerging Technological
Regimes @ https://www.triplehelixassociation.org/wp-
content/uploads/2015/09/VAGANOVA-E.-Paper-1-TRIPLE-HELIX-2015.pdf
Walton, Timothy A. Securing the Third Offset Strategy Joint Forces Quarterly 83,
3rd Quarter 2016.
Weisgerber, Marcus. The Pentagons New Algorithmic Warfare Cell Gets Its First
Mission: Hunt ISIS Defense One, 14 May 2017.
Wiley OnLine Library. Complexity, TightCoupling and Reliability: Connecting
Normal Accidents Theory and High Reliability Theory @
http://onlinelibrary.wiley.com/doi/10.1111/1468-5973.00033/abstract
Winkler, Rolfe. Elon Musk Lays Out Plans to Meld Brains and Computers The
Wall Street Journal, 20 April 2017.

99